README.md · Qwen3.6-27B

1

---

2

library_name: transformers

3

license: apache-2.0

4

license_link: https://huggingface.co/Qwen/Qwen3.6-27B/blob/main/LICENSE

5

pipeline_tag: image-text-to-text

6

---

7

8

# Qwen3.6-27B

9

10

<img width="400px" src="https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.6/logo.png">

11

12

[![Qwen Chat](https://img.shields.io/badge/%F0%9F%92%9C%EF%B8%8F%20Qwen%20Chat%20-536af5)](https://chat.qwen.ai)

13

14

> [!Note]

15

> This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format.

16

>

17

> These artifacts are compatible with Hugging Face Transformers, vLLM, SGLang, KTransformers, etc.

18

19

Following the February release of the Qwen3.5 series, we're pleased to share the first open-weight variant of Qwen3.6. Built on direct feedback from the community, Qwen3.6 prioritizes stability and real-world utility, offering developers a more intuitive, responsive, and genuinely productive coding experience.

20

21

## Qwen3.6 Highlights

22

23

This release delivers substantial upgrades, particularly in

24

25

- **Agentic Coding:** the model now handles frontend workflows and repository-level reasoning with greater fluency and precision.

26

- **Thinking Preservation:** we've introduced a new option to retain reasoning context from historical messages, streamlining iterative development and reducing overhead.

27

28

![Benchmark Results](https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3.6/Figures/qwen3.6_27b_score.png)

29

30

For more details, please refer to our blog post [Qwen3.6-27B](https://qwen.ai/blog?id=qwen3.6-27b).

31

32

## Model Overview

33

34

- Type: Causal Language Model with Vision Encoder

35

- Training Stage: Pre-training & Post-training

36

- Language Model

37

- Number of Parameters: 27B

38

- Hidden Dimension: 5120

39

- Token Embedding: 248320 (Padded)

40

- Number of Layers: 64

41

- Hidden Layout: 16 × (3 × (Gated DeltaNet → FFN) → 1 × (Gated Attention → FFN))

42

- Gated DeltaNet:

43

- Number of Linear Attention Heads: 48 for V and 16 for QK

44

- Head Dimension: 128

45

- Gated Attention:

46

- Number of Attention Heads: 24 for Q and 4 for KV

47

- Head Dimension: 256

48

- Rotary Position Embedding Dimension: 64

49

- Feed Forward Network:

50

- Intermediate Dimension: 17408

51

- LM Output: 248320 (Padded)

52

- MTP: trained with multi-steps

53

- Context Length: 262,144 natively and extensible up to 1,010,000 tokens.

54

55

56

## Benchmark Results

57

58

### Language

59

60

<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;max-width:1000px;margin:0 auto;padding:16px 0">

61

<table style="width:100%;border-collapse:collapse;font-size:13px">

62

<thead><tr>

63

<th style="padding:10px 7px;text-align:left;font-weight:600;border-bottom:2px solid #7c3aed;color:#7c3aed"></th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.5-27B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.5-397B-A17B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Gemma4-31B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Claude 4.5 Opus</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.6-35B-A3B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.6-27B</th></tr></thead>

64

<tbody>

65

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">Coding Agent</td></tr>

66

<tr>

67

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SWE-bench Verified</td>

68

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">75.0</td>

69

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">76.2</td>

70

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.0</td>

71

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.9</td>

72

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">73.4</td>

73

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">77.2</td>

74

</tr>

75

<tr>

76

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SWE-bench Pro</td>

77

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">51.2</td>

78

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">50.9</td>

79

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">35.7</td>

80

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">57.1</td>

81

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">49.5</td>

82

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">53.5</td>

83

</tr>

84

<tr>

85

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SWE-bench Multilingual</td>

86

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">69.3</td>

87

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">69.3</td>

88

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">51.7</td>

89

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">77.5</td>

90

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.2</td>

91

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">71.3</td>

92

</tr>

93

<tr>

94

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">Terminal-Bench 2.0</td>

95

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">41.6</td>

96

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.5</td>

97

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">42.9</td>

98

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">59.3</td>

99

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">51.5</td>

100

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">59.3</td>

101

</tr>

102

<tr>

103

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SkillsBench <sub><small>Avg5</small></sub></td>

104

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">27.2</td>

105

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">30.0</td>

106

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">23.6</td>

107

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">45.3</td>

108

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">28.7</td>

109

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">48.2</td>

110

</tr>

111

<tr>

112

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">QwenWebBench</td>

113

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">1068</td>

114

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">1186</td>

115

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">1197</td>

116

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">1536</td>

117

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">1397</td>

118

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">1487</td>

119

</tr>

120

<tr>

121

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">NL2Repo</td>

122

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">27.3</td>

123

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">32.2</td>

124

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">15.5</td>

125

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">43.2</td>

126

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">29.4</td>

127

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">36.2</td>

128

</tr>

129

<tr>

130

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">Claw-Eval <sub><small>Avg</small></sub></td>

131

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">64.3</td>

132

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">70.7</td>

133

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">48.5</td>

134

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">76.6</td>

135

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">68.7</td>

136

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">72.4</td>

137

</tr>

138

<tr>

139

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">Claw-Eval <sub><small>Pass^3</small></sub></td>

140

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">46.2</td>

141

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">48.1</td>

142

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">25.0</td>

143

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">59.6</td>

144

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">50.0</td>

145

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">60.6</td>

146

</tr>

147

<tr>

148

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">QwenClawBench</td>

149

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.2</td>

150

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">51.8</td>

151

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">41.7</td>

152

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.3</td>

153

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.6</td>

154

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">53.4</td>

155

</tr>

156

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">Knowledge</td></tr>

157

<tr>

158

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MMLU-Pro</td>

159

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.1</td>

160

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.8</td>

161

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.2</td>

162

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">89.5</td>

163

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.2</td>

164

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.2</td>

165

</tr>

166

<tr>

167

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MMLU-Redux</td>

168

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.2</td>

169

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">94.9</td>

170

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.7</td>

171

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">95.6</td>

172

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.3</td>

173

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.5</td>

174

</tr>

175

<tr>

176

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SuperGPQA</td>

177

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">65.6</td>

178

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">70.4</td>

179

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">65.7</td>

180

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">70.6</td>

181

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">64.7</td>

182

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">66.0</td>

183

</tr>

184

<tr>

185

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">C-Eval</td>

186

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.5</td>

187

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.0</td>

188

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">82.6</td>

189

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.2</td>

190

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.0</td>

191

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">91.4</td>

192

</tr>

193

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">STEM & Reasoning</td></tr>

194

<tr>

195

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">GPQA Diamond</td>

196

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.5</td>

197

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">88.4</td>

198

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.3</td>

199

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.0</td>

200

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.0</td>

201

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.8</td>

202

</tr>

203

<tr>

204

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">HLE</td>

205

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">24.3</td>

206

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">28.7</td>

207

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">19.5</td>

208

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">30.8</td>

209

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">21.4</td>

210

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">24.0</td>

211

</tr>

212

<tr>

213

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">LiveCodeBench v6</td>

214

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.7</td>

215

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">83.6</td>

216

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.0</td>

217

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.8</td>

218

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.4</td>

219

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">83.9</td>

220

</tr>

221

<tr>

222

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">HMMT Feb 25</td>

223

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.0</td>

224

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">94.8</td>

225

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">88.7</td>

226

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.9</td>

227

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.7</td>

228

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.8</td>

229

</tr>

230

<tr>

231

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">HMMT Nov 25</td>

232

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">89.8</td>

233

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.7</td>

234

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.5</td>

235

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.3</td>

236

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">89.1</td>

237

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.7</td>

238

</tr>

239

<tr>

240

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">HMMT Feb 26</td>

241

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.3</td>

242

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.9</td>

243

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">77.2</td>

244

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.3</td>

245

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">83.6</td>

246

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.3</td>

247

</tr>

248

<tr>

249

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">IMOAnswerBench</td>

250

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">79.9</td>

251

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.9</td>

252

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">74.5</td>

253

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.0</td>

254

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">78.9</td>

255

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.8</td>

256

</tr>

257

<tr>

258

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">AIME26</td>

259

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.6</td>

260

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.3</td>

261

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">89.2</td>

262

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">95.1</td>

263

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.7</td>

264

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">94.1</td>

265

</tr>

266

</tbody>

267

</table>

268

269

<p style="margin-top:12px;font-size:10px;opacity:0.7">

270

* SWE-Bench Series: Internal agent scaffold (bash + file-edit tools); temp=1.0, top_p=0.95, 200K context window. We correct some problematic tasks in the public set of SWE-bench Pro and evaluate all baselines on the refined benchmark.<br/>

271

* Terminal-Bench 2.0: Harbor/Terminus-2 harness; 3h timeout, 32 CPU/48 GB RAM; temp=1.0, top_p=0.95, top_k=20, max_tokens=80K, 256K ctx; avg of 5 runs.<br/>

272

* SkillsBench: Evaluated via OpenCode on 78 tasks (self-contained subset, excluding API-dependent tasks); avg of 5 runs.<br/>

273

* NL2Repo: Others are evaluated via Claude Code (temp=1.0, top_p=0.95, max_turns=900).<br/>

274

* QwenClawBench: A real-user-distribution Claw agent benchmark; temp=0.6, 256K ctx.<br/>

275

* QwenWebBench: An internal front-end code generation benchmark; bilingual (EN/CN), 7 categories (Web Design, Web Apps, Games, SVG, Data Visualization, Animation, and 3D); auto-render + multimodal judge (code/visual correctness); BT/Elo rating system.<br/>

276

* AIME 26: We use the full AIME 2026 (I & II), where the scores may differ from Qwen 3.5 notes.

277

</p>

278

279

</div>

280

281

282

### Vision Language

283

284

<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;max-width:1000px;margin:0 auto;padding:16px 0">

285

<table style="width:100%;border-collapse:collapse;font-size:13px">

286

<thead><tr>

287

<th style="padding:10px 7px;text-align:left;font-weight:600;border-bottom:2px solid #7c3aed;color:#7c3aed"></th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.5-27B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.5-397B-A17B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Gemma4-31B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Claude 4.5 Opus</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.6-35B-A3B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #7c3aed;color:#7c3aed;font-size: 14px;">Qwen3.6-27B</th></tr></thead>

288

<tbody>

289

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">STEM & Puzzle</td></tr>

290

<tr>

291

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MMMU</td>

292

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">82.3</td>

293

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.0</td>

294

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.4</td>

295

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.7</td>

296

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.7</td>

297

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">82.9</td>

298

</tr>

299

<tr>

300

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MMMU-Pro</td>

301

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">75.0</td>

302

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">79.0</td>

303

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">76.9</td>

304

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">70.6</td>

305

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">75.3</td>

306

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">75.8</td>

307

</tr>

308

<tr>

309

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MathVista <sub><small>mini</small></sub></td>

310

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.8</td>

311

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

312

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">79.3</td>

313

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

314

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.4</td>

315

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.4</td>

316

</tr>

317

<tr>

318

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">DynaMath</td>

319

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.7</td>

320

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.3</td>

321

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">79.5</td>

322

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">79.7</td>

323

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">82.8</td>

324

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.6</td>

325

</tr>

326

<tr>

327

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">VlmsAreBlind</td>

328

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">96.9</td>

329

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

330

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.2</td>

331

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

332

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">96.6</td>

333

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">97.0</td>

334

</tr>

335

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">General VQA</td></tr>

336

<tr>

337

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">RealWorldQA</td>

338

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">83.7</td>

339

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">83.9</td>

340

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">72.3</td>

341

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">77.0</td>

342

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.3</td>

343

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.1</td>

344

</tr>

345

<tr>

346

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MMStar</td>

347

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.0</td>

348

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">83.8</td>

349

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">77.3</td>

350

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">73.2</td>

351

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.7</td>

352

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.4</td>

353

</tr>

354

<tr>

355

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MMBench<sub><small>EN-DEV-v1.1</small></sub></td>

356

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.6</td>

357

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

358

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.9</td>

359

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

360

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.8</td>

361

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.3</td>

362

</tr>

363

<tr>

364

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SimpleVQA</td>

365

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">56.0</td>

366

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.1</td>

367

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.9</td>

368

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">65.7</td>

369

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">58.9</td>

370

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">56.1</td>

371

</tr>

372

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">Document Understanding</td></tr>

373

<tr>

374

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">CharXiv <sub><small>RQ</small></sub></td>

375

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">79.5</td>

376

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">80.8</td>

377

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.9</td>

378

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">68.5</td>

379

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">78.0</td>

380

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">78.4</td>

381

</tr>

382

<tr>

383

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">CC-OCR</td>

384

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.0</td>

385

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">82.0</td>

386

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">75.7</td>

387

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">76.9</td>

388

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.9</td>

389

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.2</td>

390

</tr>

391

<tr>

392

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">OCRBench</td>

393

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">89.4</td>

394

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

395

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.1</td>

396

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

397

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.0</td>

398

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">89.4</td>

399

</tr>

400

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">Spatial Intelligence</td></tr>

401

<tr>

402

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">ERQA</td>

403

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">60.5</td>

404

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.5</td>

405

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">57.5</td>

406

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">46.8</td>

407

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">61.8</td>

408

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">62.5</td>

409

</tr>

410

<tr>

411

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">CountBench</td>

412

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">97.8</td>

413

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">97.2</td>

414

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">96.1</td>

415

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.6</td>

416

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">96.1</td>

417

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">97.8</td>

418

</tr>

419

<tr>

420

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">RefCOCO <sub><small>avg</small></sub></td>

421

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.9</td>

422

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.3</td>

423

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

424

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

425

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.0</td>

426

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">92.5</td>

427

</tr>

428

<tr>

429

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">EmbSpatialBench</td>

430

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.5</td>

431

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

432

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

433

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

434

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.3</td>

435

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.6</td>

436

</tr>

437

<tr>

438

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">RefSpatialBench</td>

439

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.7</td>

440

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

441

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">4.7</td>

442

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

443

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">64.3</td>

444

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">70.0</td>

445

</tr>

446

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">Video Understanding</td></tr>

447

<tr>

448

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">VideoMME<sub><small>(w sub.)</sub></small></td>

449

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.0</td>

450

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.5</td>

451

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

452

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">77.7</td>

453

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.6</td>

454

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">87.7</td>

455

</tr>

456

<tr>

457

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">VideoMMMU</td>

458

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">82.3</td>

459

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.7</td>

460

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.6</td>

461

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.4</td>

462

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">83.7</td>

463

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.4</td>

464

</tr>

465

<tr>

466

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MLVU</td>

467

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">85.9</td>

468

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.7</td>

469

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

470

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.7</td>

471

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.2</td>

472

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86.6</td>

473

</tr>

474

<tr>

475

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MVBench</td>

476

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">74.6</td>

477

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">77.6</td>

478

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

479

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.2</td>

480

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">74.6</td>

481

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">75.5</td>

482

</tr>

483

<tr><td colspan="7" style="padding:8px 12px;font-weight:600;color:#7c3aed;border-bottom:1px solid rgba(124, 58, 237, 0.2);background:rgba(124, 58, 237, 0.1)">Visual Agent</td></tr>

484

<tr>

485

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">V*</td>

486

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">93.7</td>

487

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">95.8</td>

488

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

489

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.0</td>

490

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">90.1</td>

491

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">94.7</td>

492

</tr>

493

<tr>

494

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">AndroidWorld</td>

495

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">64.2</td>

496

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

497

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

498

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

499

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">--</td>

500

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">70.3</td>

501

</tr>

502

</tbody>

503

</table>

504

505

<p style="margin-top:12px;font-size:10px;opacity:0.7">

506

* Empty cells (--) indicate scores not yet available or not applicable.

507

</p>

508

509

</div>

510

511

512

## Quickstart

513

514

For streamlined integration, we recommend using Qwen3.6 via APIs. Below is a guide to use Qwen3.6 via OpenAI-compatible API.

515

516

### Serving Qwen3.6

517

518

Qwen3.6 can be served via APIs with popular inference frameworks.

519

In the following, we show example commands to launch OpenAI-Compatible API servers for Qwen3.6 models.

520

521

> [!Important]

522

> Inference efficiency and throughput vary significantly across frameworks.

523

> We recommend using the latest framework versions to ensure optimal performance and compatibility.

524

> For production workloads or high-throughput scenarios, dedicated serving engines such as SGLang, KTransformers or vLLM are strongly recommended.

525

526

> [!Important]

527

> The model has a default context length of 262,144 tokens.

528

> If you encounter out-of-memory (OOM) errors, consider reducing the context window.

529

> However, because Qwen3.6 leverages extended context for complex tasks, we advise maintaining a context length of at least 128K tokens to preserve thinking capabilities.

530

531

#### SGLang

532

533

[SGLang](https://github.com/sgl-project/sglang) is a fast serving framework for large language models and vision language models.

534

`sglang>=0.5.10` is recommended for Qwen3.6, which can be installed using the following command in a fresh environment:

535

```shell

536

uv pip install sglang[all]

537

```

538

See [its documentation](https://docs.sglang.ai/get_started/install.html) for more details.

539

540

The following will create API endpoints at `http://localhost:8000/v1`:

541

542

- **Standard Version**: The following command can be used to create an API endpoint with maximum context length 262,144 tokens using tensor parallel on 8 GPUs.

543

544

```shell

545

    python -m sglang.launch_server --model-path Qwen/Qwen3.6-27B --port 8000 --tp-size 8 --mem-fraction-static 0.8 --context-length 262144 --reasoning-parser qwen3

546

```

547

548

- **Tool Use**: To support tool use, you can use the following command.

549

550

```shell

551

    python -m sglang.launch_server --model-path Qwen/Qwen3.6-27B --port 8000 --tp-size 8 --mem-fraction-static 0.8 --context-length 262144 --reasoning-parser qwen3 --tool-call-parser qwen3_coder

552

```

553

554

- **Multi-Token Prediction (MTP)**: The following command is recommended for MTP:

555

556

```shell

557

    python -m sglang.launch_server --model-path Qwen/Qwen3.6-27B --port 8000 --tp-size 8 --mem-fraction-static 0.8 --context-length 262144 --reasoning-parser qwen3 --speculative-algo NEXTN --speculative-num-steps 3 --speculative-eagle-topk 1 --speculative-num-draft-tokens 4

558

```

559

560

For detailed deployment guide, see the [SGLang Qwen3.5 Cookbook](https://lmsysorg.mintlify.app/cookbook/llm/Qwen/Qwen3.5).

561

562

#### vLLM

563

564

[vLLM](https://github.com/vllm-project/vllm) is a high-throughput and memory-efficient inference and serving engine for LLMs.

565

`vllm>=0.19.0` is recommended for Qwen3.6, which can be installed using the following command in a fresh environment:

566

```shell

567

uv pip install vllm --torch-backend=auto

568

```

569

See [its documentation](https://docs.vllm.ai/en/stable/getting_started/installation/index.html) for more details.

570

571

572

The following will create API endpoints at `http://localhost:8000/v1`:

573

574

- **Standard Version**: The following command can be used to create an API endpoint with maximum context length 262,144 tokens using tensor parallel on 8 GPUs.

575

576

```shell

577

vllm serve Qwen/Qwen3.6-27B --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3

578

```

579

580

- **Tool Call**: To support tool use, you can use the following command.

581

582

```shell

583

    vllm serve Qwen/Qwen3.6-27B --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder

584

```

585

586

- **Multi-Token Prediction (MTP)**: The following command is recommended for MTP:

587

588

```shell

589

    vllm serve Qwen/Qwen3.6-27B --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --speculative-config '{"method":"qwen3_next_mtp","num_speculative_tokens":2}'

590

```

591

592

- **Text-Only**: The following command skips the vision encoder and multimodal profiling to free up memory for additional KV cache:

593

594

```shell

595

    vllm serve Qwen/Qwen3.6-27B --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --language-model-only

596

```

597

598

For detailed deployment guide, see the [vLLM Qwen3.5 Recipe](https://docs.vllm.ai/projects/recipes/en/latest/Qwen/Qwen3.5.html).

599

600

#### KTransformers

601

602

[KTransformers](https://github.com/kvcache-ai/ktransformers) is a flexible framework for experiencing cutting-edge LLM inference optimizations with CPU-GPU heterogeneous computing.

603

For running Qwen3.6 with KTransformers, see the [KTransformers Deployment Guide](https://github.com/kvcache-ai/ktransformers/blob/main/doc/en/Qwen3.5.md).

604

605

#### Hugging Face Transformers

606

607

Hugging Face Transformers contains a _lightweight_ server which can be used for quick testing and moderate load deployment.

608

The latest `transformers` is required for Qwen3.6:

609

```shell

610

pip install "transformers[serving]"

611

```

612

See [its documentation](https://huggingface.co/docs/transformers/main/serving) for more details. Please also make sure torchvision and pillow are installed.

613

614

Then, run `transformers serve` to launch a server with API endpoints at `http://localhost:8000/v1`; it will place the model on accelerators if available:

615

```shell

616

transformers serve Qwen/Qwen3.6-27B --port 8000 --continuous-batching

617

```

618

619

### Using Qwen3.6 via the Chat Completions API

620

621

The chat completions API is accessible via standard HTTP requests or OpenAI SDKs.

622

Here, we show examples using the OpenAI Python SDK.

623

624

Before starting, make sure it is installed and the API key and the API base URL is configured, e.g.:

625

```shell

626

pip install -U openai

627

628

# Set the following accordingly

629

export OPENAI_BASE_URL="http://localhost:8000/v1"

630

export OPENAI_API_KEY="EMPTY"

631

```

632

633

> [!Tip]

634

> We recommend using the following set of sampling parameters for generation

635

> - Thinking mode for general tasks: `temperature=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0`

636

> - Thinking mode for precise coding tasks (e.g. WebDev): `temperature=0.6, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0`

637

> - Instruct (or non-thinking) mode: `temperature=0.7, top_p=0.80, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0`

638

>

639

> Please note that the support for sampling parameters varies according to inference frameworks.

640

641

> [!Important]

642

> Qwen3.6 models operate in thinking mode by default, generating thinking content signified by `<think>\n...</think>\n\n` before producing the final responses.

643

> To disable thinking content and obtain direct response, refer to the examples [here](#instruct-or-non-thinking-mode).

644

645

646

#### Text-Only Input

647

648

```python

649

from openai import OpenAI

650

# Configured by environment variables

651

client = OpenAI()

652

653

messages = [

654

{"role": "user", "content": "Type \"I love Qwen3.6\" backwards"},

655

]

656

657

chat_response = client.chat.completions.create(

658

model="Qwen/Qwen3.6-27B",

659

messages=messages,

660

max_tokens=81920,

661

temperature=1.0,

662

top_p=0.95,

663

presence_penalty=0.0,

664

extra_body={

665

"top_k": 20,

666

},

667

)

668

print("Chat response:", chat_response)

669

```

670

671

672

#### Image Input

673

674

```python

675

from openai import OpenAI

676

# Configured by environment variables

677

client = OpenAI()

678

679

messages = [

680

{

681

"role": "user",

682

"content": [

683

{

684

"type": "image_url",

685

"image_url": {

686

"url": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.5/demo/CI_Demo/mathv-1327.jpg"

687

}

688

},

689

{

690

"type": "text",

691

                "text": "The centres of the four illustrated circles are in the corners of the square. The two big circles touch each other and also the two little circles. With which factor do you have to multiply the radii of the little circles to obtain the radius of the big circles?\nChoices:\n(A) $\\frac{2}{9}$\n(B) $\\sqrt{5}$\n(C) $0.8 \\cdot \\pi$\n(D) 2.5\n(E) $1+\\sqrt{2}$"

692

}

693

]

694

}

695

]

696

697

chat_response = client.chat.completions.create(

698

model="Qwen/Qwen3.6-27B",

699

messages=messages,

700

max_tokens=81920,

701

temperature=1.0,

702

top_p=0.95,

703

presence_penalty=0.0,

704

extra_body={

705

"top_k": 20,

706

},

707

)

708

print("Chat response:", chat_response)

709

```

710

711

#### Video Input

712

713

```python

714

from openai import OpenAI

715

# Configured by environment variables

716

client = OpenAI()

717

718

messages = [

719

{

720

"role": "user",

721

"content": [

722

{

723

"type": "video_url",

724

"video_url": {

725

"url": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.5/demo/video/N1cdUjctpG8.mp4"

726

}

727

},

728

{

729

"type": "text",

730

                "text": "How many porcelain jars were discovered in the niches located in the primary chamber of the tomb?"

731

}

732

]

733

}

734

]

735

736

# When vLLM is launched with `--media-io-kwargs '{"video": {"num_frames": -1}}'`,

737

# video frame sampling can be configured via `extra_body` (e.g., by setting `fps`).

738

# This feature is currently supported only in vLLM.

739

#

740

# By default, `fps=2` and `do_sample_frames=True`.

741

# With `do_sample_frames=True`, you can customize the `fps` value to set your desired video sampling rate.

742

chat_response = client.chat.completions.create(

743

model="Qwen/Qwen3.6-27B",

744

messages=messages,

745

max_tokens=81920,

746

temperature=1.0,

747

top_p=0.95,

748

presence_penalty=0.0,

749

extra_body={

750

"top_k": 20,

751

"mm_processor_kwargs": {"fps": 2, "do_sample_frames": True},

752

},

753

)

754

755

print("Chat response:", chat_response)

756

```

757

758

759

#### Instruct (or Non-Thinking) Mode

760

761

> [!Important]

762

> Qwen3.6 does not officially support the soft switch of Qwen3, i.e., `/think` and `/nothink`.

763

764

Qwen3.6 will think by default before response.

765

You can obtain direct response from the model without thinking by configuring the API parameters.

766

For example,

767

```python

768

from openai import OpenAI

769

# Configured by environment variables

770

client = OpenAI()

771

772

messages = [

773

{

774

"role": "user",

775

"content": [

776

{

777

"type": "image_url",

778

"image_url": {

779

"url": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.6/demo/RealWorld/RealWorld-04.png"

780

}

781

},

782

{

783

"type": "text",

784

"text": "Where is this?"

785

}

786

]

787

}

788

]

789

790

chat_response = client.chat.completions.create(

791

model="Qwen/Qwen3.6-27B",

792

messages=messages,

793

max_tokens=32768,

794

temperature=0.7,

795

top_p=0.8,

796

presence_penalty=1.5,

797

extra_body={

798

"top_k": 20,

799

"chat_template_kwargs": {"enable_thinking": False},

800

},

801

)

802

print("Chat response:", chat_response)

803

```

804

805

> [!Note]

806

> If you are using APIs from Alibaba Cloud Model Studio, in addition to changing `model`, please use `"enable_thinking": False` instead of `"chat_template_kwargs": {"enable_thinking": False}`.

807

808

#### Preserve Thinking

809

810

By default, only the thinking blocks generated in handling the latest user message is retained, resulting in a pattern commonly as interleaved thinking.

811

Qwen3.6 has been additionally trained to preserve and leverage thinking traces from historical messages.

812

You can enable this behavior by setting the `preserve_thinking` option:

813

```python

814

from openai import OpenAI

815

# Configured by environment variables

816

client = OpenAI()

817

818

messages = [...]

819

820

chat_response = client.chat.completions.create(

821

model="Qwen/Qwen3.6-27B",

822

messages=messages,

823

max_tokens=32768,

824

temperature=0.6,

825

top_p=0.95,

826

presence_penalty=0.0,

827

extra_body={

828

"top_k": 20,

829

"chat_template_kwargs": {"preserve_thinking": True},

830

},

831

)

832

print("Chat response:", chat_response)

833

```

834

835

> [!Note]

836

> If you are using APIs from Alibaba Cloud Model Studio, in addition to changing `model`, please use `"preserve_thinking": True` instead of `"chat_template_kwargs": {"preserve_thinking": False}`.

837

838

839

This capability is particularly beneficial for agent scenarios, where maintaining full reasoning context can enhance decision consistency and, in many cases, reduce overall token consumption by minimizing redundant reasoning. Additionally, it can improve KV cache utilization, optimizing inference efficiency in both thinking and non-thinking modes.

840

841

842

## Agentic Usage

843

844

Qwen3.6 excels in tool calling capabilities.

845

846

### Qwen-Agent

847

848

We recommend using [Qwen-Agent](https://github.com/QwenLM/Qwen-Agent) to quickly build Agent applications with Qwen3.6.

849

850

To define the available tools, you can use the MCP configuration file, use the integrated tool of Qwen-Agent, or integrate other tools by yourself.

851

```python

852

import os

853

from qwen_agent.agents import Assistant

854

855

# Define LLM

856

# Using Alibaba Cloud Model Studio

857

llm_cfg = {

858

# Use the OpenAI-compatible model service provided by DashScope:

859

'model': 'qwen3.6-27b',

860

'model_type': 'qwenvl_oai',

861

'model_server': 'https://dashscope.aliyuncs.com/compatible-mode/v1',

862

'api_key': os.getenv('DASHSCOPE_API_KEY'),

863

864

'generate_cfg': {

865

'use_raw_api': True,

866

# When using Dash Scope OAI API, pass the parameter of whether to enable thinking mode in this way

867

'extra_body': {

868

'enable_thinking': True,

869

'preserve_thinking': True,

870

},

871

},

872

}

873

874

# Using OpenAI-compatible API endpoint.

875

# functionality of the deployment frameworks and let Qwen-Agent automate the related operations.

876

#

877

# llm_cfg = {

878

# # Use your own model service compatible with OpenAI API by vLLM/SGLang:

879

# 'model': 'Qwen/Qwen3.6-27B',

880

# 'model_type': 'qwenvl_oai',

881

# 'model_server': 'http://localhost:8000/v1', # api_base

882

# 'api_key': 'EMPTY',

883

#

884

# 'generate_cfg': {

885

# 'use_raw_api': True,

886

# # When using vLLM/SGLang OAI API, pass the parameter of whether to enable thinking mode in this way

887

# 'extra_body': {

888

# 'chat_template_kwargs': {'enable_thinking': True, 'preserve_thinking': True}

889

# },

890

# },

891

# }

892

893

# Define Tools

894

tools = [

895

{'mcpServers': { # You can specify the MCP configuration file

896

"filesystem": {

897

"command": "npx",

898

"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/xxxx/Desktop"]

899

}

900

}

901

}

902

]

903

904

# Define Agent

905

bot = Assistant(llm=llm_cfg, function_list=tools)

906

907

# Streaming generation

908

messages = [{'role': 'user', 'content': 'Help me organize my desktop.'}]

909

for responses in bot.run(messages=messages):

910

pass

911

print(responses)

912

913

# Streaming generation

914

messages = [{'role': 'user', 'content': 'Develop a dog website and save it on the desktop'}]

915

for responses in bot.run(messages=messages):

916

pass

917

print(responses)

918

```

919

920

### Qwen Code

921

922

923

[Qwen Code](https://github.com/QwenLM/qwen-code) is an open-source AI agent for the terminal, optimized for Qwen models. It helps you understand large codebases, automate tedious work, and ship faster.

924

925

For more information, please refer to [Qwen Code](https://qwenlm.github.io/qwen-code-docs/).

926

927

## Processing Ultra-Long Texts

928

929

Qwen3.6 natively supports context lengths of up to 262,144 tokens.

930

For long-horizon tasks where the total length (including both input and output) exceeds this limit, we recommend using RoPE scaling techniques to handle long texts effectively., e.g., YaRN.

931

932

YaRN is currently supported by several inference frameworks, e.g., `transformers`, `vllm`, `ktransformers` and `sglang`.

933

In general, there are two approaches to enabling YaRN for supported frameworks:

934

935

- Modifying the model configuration file:

936

In the `config.json` file, change the `rope_parameters` fields in `text_config` to:

937

```json

938

{

939

"mrope_interleaved": true,

940

"mrope_section": [

941

11,

942

11,

943

10

944

],

945

"rope_type": "yarn",

946

"rope_theta": 10000000,

947

"partial_rotary_factor": 0.25,

948

"factor": 4.0,

949

"original_max_position_embeddings": 262144,

950

}

951

```

952

953

- Passing command line arguments:

954

955

For `vllm`, you can use

956

```shell

957

    VLLM_ALLOW_LONG_MAX_MODEL_LEN=1 vllm serve ... --hf-overrides '{"text_config": {"rope_parameters": {"mrope_interleaved": true, "mrope_section": [11, 11, 10], "rope_type": "yarn", "rope_theta": 10000000, "partial_rotary_factor": 0.25, "factor": 4.0, "original_max_position_embeddings": 262144}}}' --max-model-len 1010000

958

```

959

960

For `sglang` and `ktransformers`, you can use

961

```shell

962

    SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 python -m sglang.launch_server ... --json-model-override-args '{"text_config": {"rope_parameters": {"mrope_interleaved": true, "mrope_section": [11, 11, 10], "rope_type": "yarn", "rope_theta": 10000000, "partial_rotary_factor": 0.25, "factor": 4.0, "original_max_position_embeddings": 262144}}}' --context-length 1010000

963

```

964

965

> [!NOTE]

966

> All the notable open-source frameworks implement static YaRN, which means the scaling factor remains constant regardless of input length, **potentially impacting performance on shorter texts.**

967

> We advise modifying the `rope_parameters` configuration only when processing long contexts is required.

968

> It is also recommended to modify the `factor` as needed. For example, if the typical context length for your application is 524,288 tokens, it would be better to set `factor` as 2.0.

969

970

## Best Practices

971

972

To achieve optimal performance, we recommend the following settings:

973

974

1. **Sampling Parameters**:

975

- We suggest using the following sets of sampling parameters depending on the mode and task type:

976

- **Thinking mode for general tasks**:

977

`temperature=1.0`, `top_p=0.95`, `top_k=20`, `min_p=0.0`, `presence_penalty=0.0`, `repetition_penalty=1.0`

978

- **Thinking mode for precise coding tasks (e.g., WebDev)**:

979

`temperature=0.6`, `top_p=0.95`, `top_k=20`, `min_p=0.0`, `presence_penalty=0.0`, `repetition_penalty=1.0`

980

- **Instruct (or non-thinking) mode**:

981

`temperature=0.7`, `top_p=0.80`, `top_k=20`, `min_p=0.0`, `presence_penalty=1.5`, `repetition_penalty=1.0`

982

   - For supported frameworks, you can adjust the `presence_penalty` parameter between 0 and 2 to reduce endless repetitions. However, using a higher value may occasionally result in language mixing and a slight decrease in model performance.

983

984

2. **Adequate Output Length**: We recommend using an output length of 32,768 tokens for most queries. For benchmarking on highly complex problems, such as those found in math and programming competitions, we suggest setting the max output length to 81,920 tokens. This provides the model with sufficient space to generate detailed and comprehensive responses, thereby enhancing its overall performance.

985

986

3. **Standardize Output Format**: We recommend using prompts to standardize model outputs when benchmarking.

987

- **Math Problems**: Include "Please reason step by step, and put your final answer within \boxed{}." in the prompt.

988

   - **Multiple-Choice Questions**: Add the following JSON structure to the prompt to standardize responses: "Please show your choice in the `answer` field with only the choice letter, e.g., `"answer": "C"`."

989

990

4. **Long Video Understanding**: To optimize inference efficiency for plain text and images, the `size` parameter in the released `video_preprocessor_config.json` is conservatively configured. It is recommended to set the `longest_edge` parameter in the video_preprocessor_config file to 469,762,048 (corresponding to 224k video tokens) to enable higher frame-rate sampling for hour-scale videos and thereby achieve superior performance. For example,

991

```json

992

{"longest_edge": 469762048, "shortest_edge": 4096}

993

```

994

995

    Alternatively, override the default values via engine startup parameters. For implementation details, refer to: [vLLM](https://github.com/vllm-project/vllm/pull/34330) / [SGLang](https://github.com/sgl-project/sglang/pull/18467).

996

997

998

### Citation

999

1000

If you find our work helpful, feel free to give us a cite.

1001

1002

```bibtex

1003

@misc{qwen3.6-27b,

1004

title = {{Qwen3.6-27B}: Flagship-Level Coding in a {27B} Dense Model},

1005

author = {{Qwen Team}},

1006

month = {April},

1007

year = {2026},

1008

url = {https://qwen.ai/blog?id=qwen3.6-27b}

1009

}

1010

```