README.md
63.6 KB · 671 lines · markdown Raw
1 ---
2 library_name: transformers
3 license: other
4 license_name: qwen-community-1.0
5 license_link: LICENSE
6 pipeline_tag: image-text-to-text
7 ---
8
9 # Qwen3.8-Flash-Next
10
11 > [!Note]
12 > This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format.
13 >
14 > These artifacts are compatible with Hugging Face Transformers, vLLM, SGLang, TokenSpeed, etc.
15
16 > [!Tip]
17 > For users seeking managed, scalable inference without infrastructure maintenance, the official Qwen API service is provided by [Qwen Cloud](https://www.qwencloud.com).
18 >
19 > In particular, **Qwen3.8-Flash** is the official version based on Qwen3.8-Flash-Next with more production features, e.g., 1M context length by default, official built-in tools. For more information, please refer to the [Qwen3.8-Flash Overview](https://www.qwencloud.com/models/qwen3.8-flash).
20
21
22 As the frontier of foundation models pushes toward ever-larger parameter counts and ever-longer context windows, the question is no longer just how much we can scale, but how efficiently we can do so. Sustainable progress toward artificial general intelligence (AGI) that benefits everyone demands architectural innovation. Today, we are sharing a concrete step in that direction: Qwen3.8-Flash-Next.
23
24 ![Qwen3.8-Flash-Next Architecture](https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.8-Flash-Next/architecture.png)
25
26 This experimental preview of the architecture that will underpin Qwen4 is built around a fundamental rethinking of how the core components of modern large language models (LLMs) interact at scale.
27
28 ## Highlights
29
30 The first open-weight release under this architecture is Qwen3.8-Flash-Next, which introduces:
31
32 - **Hybrid Attention with QSA**: The Gated DeltaNet and Gated Attention pairing has been reworked into Gated DeltaNet and Qwen Sparse Attention (QSA). Rather than selecting individual tokens for processing, QSA operates at the micro-block level. This cuts long-context latency significantly, a critical gain as agentic workloads increasingly dominate real-world usage.
33 - **Gated Residual**: Residual streams with normalization are what make deep LLM training manageable. Gated Residual modulates information flowing through widened residual streams via an element-wise, data-dependent read gate and a per-branch scalar write gate. This brings finer-grained expressiveness across layers while preserving training stability and keeping inference overhead low.
34 - **N-gram Embedding**: Embeddings provide a unique axis for parameter scaling that requires less computation and is more amenable to offloading than Mixture-of-Experts (MoE). By indexing with short n-grams, this approach makes parameter scaling highly efficient for memory-constrained accelerators without sacrificing quality.
35 - **Tailored Training Recipe**: The Muon and AdamW optimizers are applied to specific weight categories to maximize efficiency. Guided by refitted scaling laws, we eliminate traditional batch-size warmups and start directly at the target batch size, substantially reducing total optimizer steps while safely supporting larger learning rates for robust convergence.
36
37 For more details, please refer to our blog post [Qwen3.8-Flash-Next](https://qwen.ai/blog?id=qwen3.8-flash-next) and [the technical report](https://github.com/QwenLM/Qwen3.8-Flash-Next/blob/main/tech_report.pdf).
38
39 We are excited to embark on this next chapter with you and welcome your feedback as we build what comes next.
40
41 ## Model Overview
42
43 - Type: Causal Language Model with Vision Encoder
44 - Training Stage: Pre-training & Post-training
45 - Language Model
46 - Number of Parameters: 125B with 6B activated, plus 51B n-gram embedding and 4B MTP
47 - Hidden Dimension: 2560
48 - Token Embedding: 248320 (Padded)
49 - N-gram Embedding: 20,000,000 (bigrams/trigrams at layer 2)
50 - Number of Layers: 48
51 - Hidden Layout: 12 × (3 × (Gated DeltaNet → MoE) → 1 × (Qwen Sparse Attention → MoE))
52 - Gated DeltaNet:
53 - Number of Linear Attention Heads: 48 for V and 16 for QK
54 - Head Dimension: 128
55 - Qwen Sparse Attention:
56 - Number of Attention Heads: 24 for Q and 2 for KV
57 - Head Dimension: 256
58 - Rotary Position Embedding Dimension: 64
59 - Indexer Structure: MQA with 4 Query Heads and 1 Shared Key Head
60 - Indexer Head Dimension: 128
61 - Budget: 512 blocks or 2048 tokens
62 - Mixture Of Experts
63 - Number of Experts: 512
64 - Number of Activated Experts: 10 Routed + 1 Shared
65 - Expert Intermediate Dimension: 640
66 - Gated Residual:
67 - Number of Branches: 4
68 - Bottleneck Rank: 320
69 - LM Output: 248320 (Padded)
70 - MTP: 1 layer, trained with multi-steps
71 - Context Length: 262,144 natively and extensible up to 1,000,000 tokens.
72
73 ## Benchmark Results
74
75 <style>
76 .vl-table th{font-size:15px!important;line-height:1.2;color:inherit!important}
77 .vl-table td:not(.benchmark-cell):not([colspan]){font-size:15px;line-height:1.2;vertical-align:middle;color:inherit!important}
78 .vl-table .benchmark-cell{padding:12px 10px 12px 18px!important;vertical-align:middle}
79 .vl-table .benchmark-capability{font-size:15px;font-weight:600;line-height:1.22;color:inherit!important}
80 .vl-table .benchmark-name{margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;opacity:.6}
81 .vl-table .metric-stack{display:flex;flex-direction:column;gap:7px;padding:3px 0}
82 .vl-table .metric-label{font-size:10px;font-weight:400;line-height:1.1;opacity:.55}
83 .vl-table .metric-value{margin-top:2px;font-size:15px;line-height:1.15;color:inherit!important}
84 .vl-table .metric-pair{white-space:nowrap}
85 .vl-table .metric-sep{color:inherit!important;opacity:.55;padding:0 3px}
86
87 @media (prefers-color-scheme: dark){
88 .vl-table th,.vl-table td[colspan]{color:#8DA2FF!important;border-bottom-color:#8DA2FF!important}
89 .vl-table td[colspan]{background:rgba(10,46,254,0.22)!important}
90 }
91 </style>
92
93 ### Language
94
95 <div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;max-width:1200px;margin:0 auto;padding:16px 0">
96 <table class="vl-table" style="width:100%;table-layout:fixed;border-collapse:collapse;font-size:13px">
97 <thead><tr>
98 <th style="padding:10px 7px;text-align:left;font-weight:600;border-bottom:2px solid #0A2EFE;color:#0A2EFE"></th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:14.00%;background:rgba(10, 46, 254, 0.08);">Qwen3.8-Flash-Next</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:14.00%;">Qwen3.8-27B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:14.00%;">Qwen3.7-Plus</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:14.00%;">DeepSeek-V4-Flash-0731</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:14.00%;">Claude-Opus-4.6 (Max)</th></tr></thead>
99 <tbody>
100 <tr>
101 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717"># Params</div></td>
102 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;">125B</td>
103 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">27B</td>
104 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">397B</td>
105 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">284B</td>
106 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
107 </tr>
108 <tr>
109 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717"># Activated params</div></td>
110 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;">6B</td>
111 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">27B</td>
112 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">17B</td>
113 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">13B</td>
114 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
115 </tr>
116 <tr>
117 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717"># N-gram embedding params</div></td>
118 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;">51B</td>
119 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
120 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
121 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
122 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
123 </tr>
124 <tr><td colspan="6" style="padding:8px 12px;font-weight:600;color:#0A2EFE;border-bottom:1px solid rgba(10, 46, 254, 0.2);background:#D6DAFC">Coding</td></tr>
125 <tr>
126 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Agentic coding</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">DeepSWE 1.1</div></td>
127 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>58.7</strong></td>
128 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">42.2</td>
129 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">16.5</td>
130 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">54.4</td>
131 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
132 </tr>
133 <tr>
134 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Agentic coding</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">SWE-bench Pro</div></td>
135 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>62.5</strong></td>
136 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">61.7</td>
137 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">55.8</td>
138 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">56.0</td>
139 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">53.4</td>
140 </tr>
141 <tr>
142 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Multilingual software engineering</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">SWE-bench Multilingual</div></td>
143 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>81.0</strong></td>
144 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">73.8</td>
145 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">75.8</td>
146 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
147 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">77.5</td>
148 </tr>
149 <tr>
150 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Repo-level code generation</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">NL2Repo-Bench</div></td>
151 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;">48.1</td>
152 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">42.3</td>
153 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">41.1</td>
154 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>54.2</strong></td>
155 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">47.6</td>
156 </tr>
157 <tr><td colspan="6" style="padding:8px 12px;font-weight:600;color:#0A2EFE;border-bottom:1px solid rgba(10, 46, 254, 0.2);background:#D6DAFC">Agent</td></tr>
158 <tr>
159 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Long-horizon office work</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">CoWorkBench</div></td>
160 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>73.9</strong></td>
161 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">70.7</td>
162 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">65.1</td>
163 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">45.1</td>
164 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">68.2</td>
165 </tr>
166 <tr>
167 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Professional job tasks</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">JobBench</div></td>
168 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>55.7</strong></td>
169 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">33.4</td>
170 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">27.6</td>
171 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">41.3</td>
172 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">36.6</td>
173 </tr>
174 <tr>
175 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Frontier agentic tasks</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">Agents' Last Exam</div></td>
176 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@1</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">24.3</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Score</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>51.2</strong></div></div></div></td>
177 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@1</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">20.4</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Score</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">42.9</div></div></div></td>
178 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@1</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">13.2</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Score</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">33.6</div></div></div></td>
179 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@1</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>25.2</strong></div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Score</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">--</div></div></div></td>
180 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
181 </tr>
182 <tr>
183 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Real-world tool use</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">Toolathlon Verified (Pass@1)</div></td>
184 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>73.5</strong></td>
185 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">67.1</td>
186 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">50.6</td>
187 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">70.3</td>
188 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
189 </tr>
190 <tr><td colspan="6" style="padding:8px 12px;font-weight:600;color:#0A2EFE;border-bottom:1px solid rgba(10, 46, 254, 0.2);background:#D6DAFC">General</td></tr>
191 <tr>
192 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Instruction following</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">IFBench</div></td>
193 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>81.3</strong></td>
194 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">79.5</td>
195 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">79.1</td>
196 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">79.2</td>
197 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">62.5</td>
198 </tr>
199 <tr>
200 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Scientific reasoning</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">GPQA Diamond</div></td>
201 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>91.7</strong></td>
202 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">89.2</td>
203 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">90.3</td>
204 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">90.8</td>
205 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">91.3</td>
206 </tr>
207 <tr>
208 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Multidisciplinary reasoning</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">HLE</div></td>
209 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;">35.9</td>
210 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">30.8</td>
211 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">34.7</td>
212 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">33.8</td>
213 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>40.0</strong></td>
214 </tr>
215 <tr>
216 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Competitive coding</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">LiveCodeBench v6</div></td>
217 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>91.9</strong></td>
218 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">90.3</td>
219 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">89.6</td>
220 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">90.6</td>
221 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">88.8</td>
222 </tr>
223 </tbody>
224 </table>
225 <p style="margin-top:12px;font-size:10px;line-height:1.4;opacity:0.7">1. DeepSWE 1.1: evaluated with the Claude Code and mini-SWE-agent harnesses, temp=1.0, top_p=0.95, 256K context window. We report the highest score across the two harnesses; notably, Qwen3.8-Flash-Next performs best on mini-SWE-agent.<br>2. SWE-bench Pro: except for Claude-Opus-4.6 (Max), for which we report the officially published score, all models are evaluated with the Claude Code harness, temp=1.0, top_p=0.95, 256K context window. Problematic tasks were corrected and all baseline models were re-evaluated on the refined benchmark.<br>3. SWE-bench Multilingual: evaluated with the mini-SWE-agent harness, temp=1.0, top_p=0.95, 256K context window.<br>4. NL2Repo-Bench: evaluated with the Claude Code harness. To prevent reward hacking, we disable Bash commands that attempt to access the specific repository, such as pip download, pip install and git clone.<br>5. CoWorkBench: an in-house cowork benchmark for evaluating long-horizon office and productivity agent tasks across computer science, finance, law, medical and other productivity domains.<br>6. HLE: judged by GPT-4o.<br>7. The best result in each row is shown in bold.<br>8. Empty cells (--): scores are not yet available or are not applicable.</p>
226 </div>
227
228 ### Vision Language
229
230 <div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;max-width:1000px;margin:0 auto;padding:16px 0">
231 <table class="vl-table" style="width:100%;table-layout:fixed;border-collapse:collapse;font-size:13px">
232 <thead><tr>
233 <th style="padding:10px 7px;text-align:left;font-weight:600;border-bottom:2px solid #0A2EFE;color:#0A2EFE"></th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:16.00%;background:rgba(10, 46, 254, 0.08);">Qwen3.8-Flash-Next</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:16.00%;">Qwen3.8-27B</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:16.00%;">Qwen3.7-Plus</th><th style="padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #0A2EFE;color:#0A2EFE;font-size: 14px;width:16.00%;">Claude-Opus-4.6 (Max)</th></tr></thead>
234 <tbody>
235 <tr><td colspan="5" style="padding:8px 12px;font-weight:600;color:#0A2EFE;border-bottom:1px solid rgba(10, 46, 254, 0.2);background:#D6DAFC">Agentic Multimodal Intelligence</td></tr>
236 <tr>
237 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Multimodal tool use</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">ClawEval-MM</div></td>
238 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@3</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>64.4</strong></div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Average</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>60.4</strong></div></div></div></td>
239 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@3</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">57.4</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Average</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">56.9</div></div></div></td>
240 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@3</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">57.4</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Average</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">60.1</div></div></div></td>
241 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Pass@3</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">52.5</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Average</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">54.7</div></div></div></td>
242 </tr>
243 <tr>
244 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Application recreation</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">RecreationBench</div></td>
245 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>49.9</strong></td>
246 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">47.1</td>
247 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">30.2</td>
248 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
249 </tr>
250 <tr>
251 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Mobile use</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">AndroidWorld</div></td>
252 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>84.5</strong></td>
253 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">81.9</td>
254 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">81.0</td>
255 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">62.0</td>
256 </tr>
257 <tr>
258 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Computer use</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">OSWorld 2.0</div></td>
259 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Binary</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>19.4</strong></div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Partial</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>52.3</strong></div></div></div></td>
260 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Binary</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">19.4</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Partial</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">48.0</div></div></div></td>
261 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Binary</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">2.8</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Partial</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">21.5</div></div></div></td>
262 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
263 </tr>
264 <tr>
265 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Visual web development</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">Vision2Web</div></td>
266 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>64.0</strong></td>
267 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">62.9</td>
268 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">42.1</td>
269 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">--</td>
270 </tr>
271 <tr><td colspan="5" style="padding:8px 12px;font-weight:600;color:#0A2EFE;border-bottom:1px solid rgba(10, 46, 254, 0.2);background:#D6DAFC">General Multimodal Intelligence</td></tr>
272 <tr>
273 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Embodied intelligence</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">ERQA</div></td>
274 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>72.3</strong></td>
275 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">65.5</td>
276 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">69.8</td>
277 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">40.8</td>
278 </tr>
279 <tr>
280 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Long video understanding</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">LVBench</div></td>
281 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>76.6</strong></td>
282 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">72.4</td>
283 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">76.2</td>
284 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">63.0</td>
285 </tr>
286 <tr>
287 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Real-world perception</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">RealWorldQA</div></td>
288 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><strong>88.5</strong></td>
289 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">85.9</td>
290 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">86.9</td>
291 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;">73.9</td>
292 </tr>
293 <tr>
294 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Visual math problem solving</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">MathVision</div></td>
295 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>90.6</strong></div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">With CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>95.7</strong></div></div></div></td>
296 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">90.0</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">With CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">94.6</div></div></div></td>
297 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">90.3</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">With CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">88.7</div></div></div></td>
298 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">65.5</div></div></div></td>
299 </tr>
300 <tr>
301 <td class="benchmark-cell" style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);"><div class="benchmark-capability" style="font-size:15px;font-weight:600;line-height:1.22;color:#171717">Scientific chart analysis</div><div class="benchmark-name" style="margin-top:4px;font-size:11px;font-weight:400;line-height:1.2;color:#6B6B6B">CharXiv (RQ)</div></td>
302 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);background:rgba(10, 46, 254, 0.08);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">84.6</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">With CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>90.6</strong></div></div></div></td>
303 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">83.7</div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">With CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">90.2</div></div></div></td>
304 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717"><strong>85.8</strong></div></div><div style="margin-top:7px"><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">With CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">85.9</div></div></div></td>
305 <td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);vertical-align:middle;font-size:15px;line-height:1.2;"><div class="metric-stack" style="padding:3px 0"><div><div class="metric-label" style="font-size:10px;font-weight:400;line-height:1.1;color:#777">Without CI</div><div class="metric-value" style="margin-top:2px;font-size:15px;line-height:1.15;color:#171717">66.0</div></div></div></td>
306 </tr>
307 </tbody>
308 </table>
309 <p style="margin-top:12px;font-size:10px;line-height:1.4;opacity:0.7">1. ClawEval-MM: scores are reported as "pass@3 / average score". Pass@3 measures the percentage passed in at least one of three trials, and the average score is the mean score across the three trials.<br>2. RecreationBench: an in-house long-horizon application-recreation benchmark for evaluating hybrid-agent abilities spanning five platforms — desktop (Ubuntu, macOS, Windows), mobile (Android) and web.<br>3. OSWorld 2.0: scores are reported as "binary / partial". The binary score is the percentage of tasks that receive the full task reward, while the partial score aggregates the partial rewards obtained across all tasks.<br>4. Vision2Web: scores are reported as the average over the frontend, webpage and website categories, using the Claude Code harness and judged by gpt-5.4-2026-03-05.<br>5. MathVision, CharXiv (RQ): scores are reported as "without CI / with CI". A small number of incorrect ground-truth annotations in MathVision were corrected after manual verification. Our model's score is evaluated using a fixed prompt, e.g. "Please reason step by step, and put your final answer within \boxed{}." For other models, we report the higher score between runs with and without the \boxed{} formatting.<br>6. The best result in each row is shown in bold.<br>7. Empty cells (--) indicate scores not yet available or not applicable.</p>
310 </div>
311
312
313 ## Quickstart
314
315 For streamlined integration, we recommend using Qwen3.8-Flash-Next via APIs.
316
317 ### Serving Qwen3.8-Flash-Next
318
319 > [!Important]
320 > Inference efficiency and throughput vary significantly across frameworks.
321 > We recommend using the latest framework versions to ensure optimal performance and compatibility.
322 > For production workloads or high-throughput scenarios, dedicated serving engines such as SGLang, KTransformers or vLLM are strongly recommended.
323
324
325 Qwen3.8-Flash-Next can be deployed with popular inference frameworks, e.g.:
326
327 - [SGLang](https://www.sglang.io/): [Qwen3.8-Flash-Next Cookbook](https://docs.sglang.io/cookbook/autoregressive/Qwen/Qwen3.8-Flash-Next)
328 - [vLLM](https://vllm.ai/): [Qwen3.8-Flash-Next Recipe](https://recipes.vllm.ai/Qwen/Qwen3.8-Flash-Next)
329 - [TokenSpeed](https://lightseek.org/tokenspeed/): [Qwen3.8-Flash-Next Recipe](https://lightseek.org/tokenspeed/recipes/models#qwen3-8-flash-next)
330
331
332 ### API Usage
333
334 > [!Important]
335 > Qwen3.8-Flash-Next models operate in thinking mode by default, generating thinking content signified by `<think>\n...</think>\n\n` before producing the final responses.
336 > To disable thinking content and obtain direct response, refer to the examples [here](#instruct-or-non-thinking-mode).
337
338 > [!Tip]
339 > We recommend using the following sets of sampling parameters for generation:
340 > - Thinking Mode: `temperature=1.0`, `top_p=0.95`, `top_k=20`, `min_p=0.0`, `presence_penalty=0.0`, `repetition_penalty=1.0`
341 > - 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`
342 >
343 > Please note that the support for sampling parameters varies according to inference frameworks.
344
345 > [!Tip]
346 > In multi-turn agentic tasks, lower reasoning effort does not always reduce overall task completion time. Although it may produce faster per-turn responses, it can also lead to insufficient analysis, more failures, and repeated retries, which may increase total latency and token consumption.
347
348 Qwen3.8-Flash-Next supports controlling thinking behavior via `enable_thinking`, `preserve_thinking`, and `reasoning_effort`.
349
350 #### Chat Completions API
351
352 The Chat Completions API can be used with most inference frameworks, as well as [Qwen Cloud](https://www.qwencloud.com/).
353 Before starting, make sure it is installed and the API key and the API base URL is configured, e.g.:
354 ```shell
355 pip install -U openai
356
357 # Set the following accordingly
358 export OPENAI_BASE_URL="http://localhost:8000/v1"
359 export OPENAI_API_KEY="EMPTY"
360 ```
361
362 ##### Text-Only Input
363
364 ```python
365 from openai import OpenAI
366 # Configured by environment variables
367 client = OpenAI()
368
369 messages = [
370 {"role": "user", "content": "Write a Python function to merge two sorted linked lists."},
371 ]
372
373 completion = client.chat.completions.create(
374 model="Qwen/Qwen3.8-Flash-Next",
375 messages=messages,
376 extra_body={
377 "chat_template_kwargs": {
378 "enable_thinking": True, # on by default
379 "preserve_thinking": True, # on by default
380 },
381 },
382 reasoning_effort="xhigh", # xhigh by default; supported levels are xhigh, medium, and low
383 stream=True,
384 stream_options={"include_usage": True},
385 )
386
387 reasoning_content = ""
388 answer_content = ""
389 is_answering = False
390 print("\n" + "=" * 20 + "Reasoning" + "=" * 20 + "\n")
391
392 for chunk in completion:
393 if not chunk.choices:
394 print("\nUsage:")
395 print(chunk.usage)
396 continue
397
398 delta = chunk.choices[0].delta
399
400 if hasattr(delta, "reasoning_content") and delta.reasoning_content is not None:
401 if not is_answering:
402 print(delta.reasoning_content, end="", flush=True)
403 reasoning_content += delta.reasoning_content
404 elif hasattr(delta, "reasoning") and delta.reasoning is not None:
405 if not is_answering:
406 print(delta.reasoning, end="", flush=True)
407 reasoning_content += delta.reasoning
408
409 if hasattr(delta, "content") and delta.content:
410 if not is_answering:
411 print("\n" + "=" * 20 + "Answer" + "=" * 20 + "\n")
412 is_answering = True
413 print(delta.content, end="", flush=True)
414 answer_content += delta.content
415
416 messages.append({
417 "role": "assistant",
418 "content": answer_content,
419 "reasoning_content": reasoning_content,
420 "reasoning": reasoning_content,
421 })
422 ```
423
424
425 ##### Image Input
426
427 ```python
428 from openai import OpenAI
429 # Configured by environment variables
430 client = OpenAI()
431
432 messages = [
433 {
434 "role": "user",
435 "content": [
436 {
437 "type": "image_url",
438 "image_url": {
439 "url": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.5/demo/CI_Demo/mathv-1327.jpg"
440 }
441 },
442 {
443 "type": "text",
444 "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) $1+\\sqrt{2}$"
445 }
446 ]
447 }
448 ]
449
450 chat_response = client.chat.completions.create(
451 model="Qwen/Qwen3.8-Flash-Next",
452 messages=messages,
453 )
454 print("Chat response:", chat_response)
455 ```
456
457 ##### Video Input
458
459 ```python
460 from openai import OpenAI
461 # Configured by environment variables
462 client = OpenAI()
463
464 messages = [
465 {
466 "role": "user",
467 "content": [
468 {
469 "type": "video_url",
470 "video_url": {
471 "url": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.5/demo/video/N1cdUjctpG8.mp4"
472 }
473 },
474 {
475 "type": "text",
476 "text": "How many porcelain jars were discovered in the niches located in the primary chamber of the tomb?"
477 }
478 ]
479 }
480 ]
481
482 chat_response = client.chat.completions.create(
483 model="Qwen/Qwen3.8-Flash-Next",
484 messages=messages,
485 )
486
487 # When vLLM is launched with `--media-io-kwargs '{"video": {"num_frames": -1}}'`,
488 # video frame sampling can be configured via `extra_body` (e.g., by setting `fps`).
489 # This feature is currently supported only in vLLM.
490 #
491 # By default, `fps=2` and `do_sample_frames=True`.
492 # With `do_sample_frames=True`, you can customize the `fps` value to set your desired video sampling rate.
493 # chat_response = client.chat.completions.create(
494 # model="Qwen/Qwen3.8-Flash-Next",
495 # messages=messages,
496 # extra_body={
497 # "mm_processor_kwargs": {"fps": 2, "do_sample_frames": True},
498 # },
499 # )
500
501 print("Chat response:", chat_response)
502 ```
503
504
505 ##### Instruct (or Non-Thinking) Mode
506
507 Qwen3.8-Flash-Next will think by default before responding.
508 You can obtain a direct response from the model without thinking by configuring the API parameters.
509 For example,
510 ```python
511 from openai import OpenAI
512 # Configured by environment variables
513 client = OpenAI()
514
515 messages = [
516 {
517 "role": "user",
518 "content": [
519 {
520 "type": "image_url",
521 "image_url": {
522 "url": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.5/demo/RealWorld/RealWorld-04.png"
523 }
524 },
525 {
526 "type": "text",
527 "text": "Where is this?"
528 }
529 ]
530 }
531 ]
532
533 chat_response = client.chat.completions.create(
534 model="Qwen/Qwen3.8-Flash-Next",
535 messages=messages,
536 temperature=0.7,
537 top_p=0.8,
538 presence_penalty=1.5,
539 extra_body={
540 "top_k": 20,
541 "chat_template_kwargs": {"enable_thinking": False},
542 },
543 )
544 print("Chat response:", chat_response)
545 ```
546
547 > [!Note]
548 > If you are using APIs from Qwen Cloud, in addition to changing `model`, please use `"enable_thinking": False` instead of `"chat_template_kwargs": {"enable_thinking": False}`.
549
550
551 ##### Disable Preserved Thinking
552
553 By default, Qwen3.8-Flash-Next retains thinking blocks from all historical messages, maintaining a complete reasoning trace across the conversation. This behavior, known as preserved thinking, ensures full context continuity and is especially beneficial for agent scenarios where decision consistency and reduced redundant reasoning are critical. It also improves KV cache utilization, optimizing inference efficiency in both thinking and non-thinking modes.
554
555 If you prefer to retain only the thinking blocks from the latest user message, you can disable this behavior by setting `preserve_thinking` to `False`:
556
557 ```python
558 from openai import OpenAI
559
560 # Configured by environment variables
561 client = OpenAI()
562 messages = [...]
563 chat_response = client.chat.completions.create(
564 model="Qwen/Qwen3.8-Flash-Next",
565 messages=messages,
566 extra_body={
567 "chat_template_kwargs": {"preserve_thinking": False},
568 },
569 )
570 print("Chat response:", chat_response)
571 ```
572
573 > [!Note]
574 > If you are using APIs from Qwen Cloud, in addition to changing `model`, please use `"preserve_thinking": False` directly instead of wrapping it in `chat_template_kwargs`.
575
576
577
578 ## Best Practices
579
580 To achieve optimal performance, we recommend the following settings:
581
582 1. **Sampling Parameters**: We suggest using the following sets of sampling parameters:
583
584 - Thinking Mode: `temperature=1.0`, `top_p=0.95`, `top_k=20`, `min_p=0.0`, `presence_penalty=0.0`, `repetition_penalty=1.0`
585 - 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`
586
587 For supported frameworks, you can adjust the `presence_penalty` parameter between 0 and 2 to reduce endless repetition. However, using a higher value may occasionally result in language mixing and a slight decrease in model performance.
588
589 2. **Adequate Output Length**: To optimize performance on agentic tasks, we recommend allocating sufficient output length to allow the model to generate detailed and comprehensive responses. For frameworks that support separate token limits for internal reasoning and final outputs, we suggest the following configuration within the 1M context length:
590
591 - Reasoning Content: Set the maximum output length to 262,144 tokens.
592 - Final Response: Set the maximum output length to 131,072 tokens.
593
594 These settings provide the necessary capacity for complex reasoning while ensuring ample space for high-quality final deliverables.
595
596 3. **Processing Ultra-Long Texts**: Qwen3.8-Flash-Next natively supports context lengths of up to 262,144 tokens. 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.
597
598 YaRN is currently supported by several inference frameworks, e.g., vLLM, SGLang, and TokenSpeed.
599 In general, there are two approaches to enabling YaRN for supported frameworks:
600
601 - Modifying the model configuration file:
602
603 In the `config.json` file, change the `rope_parameters` fields in `text_config` to:
604 ```json
605 {
606 "mrope_interleaved": true,
607 "mrope_section": [
608 11,
609 11,
610 10
611 ],
612 "rope_type": "yarn",
613 "rope_theta": 10000000,
614 "partial_rotary_factor": 0.25,
615 "factor": 4.0,
616 "original_max_position_embeddings": 262144
617 }
618 ```
619
620 - Passing command line arguments:
621
622 For vLLM, you can use
623 ```shell
624 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 1000000
625 ```
626
627 For SGLang, you can use
628 ```shell
629 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 1000000
630 ```
631
632 For TokenSpeed, you can use
633 ```shell
634 TOKENSPEED_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 tokenspeed 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 1000000
635 ```
636
637 > [!NOTE]
638 > 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.**
639 > We advise modifying the `rope_parameters` configuration only when processing long contexts is required.
640 > 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.
641
642
643 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,
644 ```json
645 {"longest_edge": 469762048, "shortest_edge": 4096}
646 ```
647
648 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).
649
650
651 ### Citation
652
653 If you find our work helpful, feel free to give us a cite.
654
655 ```bibtex
656 @techreport{qwen2026design,
657 title = {On the Design of {Qwen3.8-Next} Architecture: Evaluation, Efficiency, and Training Stability},
658 author = {{Qwen Team}},
659 institution = {Alibaba Group},
660 month = {August},
661 year = {2026}
662 }
663
664 @misc{qwen3.8flashnext,
665 title = {{Qwen3.8-Flash-Next}: A New Architecture, Towards Ultimate Cost-Efficiency},
666 author = {{Qwen Team}},
667 month = {August},
668 year = {2026},
669 url = {https://qwen.ai/blog?id=qwen3.8-flash-next}
670 }
671 ```