README.md · Ornith-1.0-9B-GGUF

1

---

2

library_name: transformers

3

license: mit

4

license_link: https://huggingface.co/deepreinforce-ai/Ornith-1.0-9B-GGUF/blob/main/LICENSE

5

pipeline_tag: text-generation

6

---

7

8

9

<img width="600px" src="assets/ornith_logo.png">

10

11

[![Ornith Blog](https://img.shields.io/badge/%F0%9F%A6%A2%EF%B8%8F%20Ornith%20Blog%20-FD8E5B)](https://deep-reinforce.com/ornith.html)

12

13

14

# Ornith-1.0-9B-GGUF

15

16

17

Aloha! 🌺 Today, we are releasing Ornith-1.0, a self-improving family of open-source models for agentic coding.

18

19

Highlights:

20

21

- **State-of-the-Art Coding Agents**: Available in 9B-Dense, 31B-Dense, 35B-MoE, and 397B-MoE (post-trained on top of Gemma 4 and Qwen 3.5), achieving state-of-the-art performance among open-source models of comparable size on coding benchmarks such as Terminal-Bench 2.1, SWE-Bench, NL2Repo and OpenClaw.

22

- **Self-Improving Training Framework**:  Ornith-1.0 employs RL to learn to generate not only solution rollouts, but also the scallfold that drive those rollouts. By jointly optimizing the scaffold and the resulting solution, the model  discovers better search trajectories and generates higher-quality solutions.

23

- **Licence**: MIT licensed, globally accessible, and free from regional limitations.

24

25

<img style="width: 100%; max-width: 900px;" src="assets/ornith_9b_eval.png" alt="Ornith 9B Benchmark Results" title="Ornith 9B Benchmark Results">

26

27

## Ornith 1.0 9B

28

29

This model card documents **Ornith-1.0-9B**, the most lightweight member of the Ornith family, designed for efficient single-GPU deployment.

30

31

### Benchmarks

32

33

34

<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;width:100%;margin:0 auto;padding:16px 0">

35

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

36

<thead><tr>

37

<th style="width:28%;padding:10px 7px;text-align:left;font-weight:600;border-bottom:2px solid #FD8E5B;color:#FD8E5B"></th>

38

<th style="width:14.40%;padding:10px 7px;text-align:center;font-weight:700;border-bottom:2px solid #FD8E5B;color:#FD8E5B;font-size:14px;background:rgba(253, 142, 91, 0.12)">Ornith-1.0-9B</th>

39

<th style="width:14.40%;padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #FD8E5B;color:#FD8E5B;font-size:14px">Qwen3.5-9B</th>

40

<th style="width:14.40%;padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #FD8E5B;color:#FD8E5B;font-size:14px">Qwen3.5-35B</th>

41

<th style="width:14.40%;padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #FD8E5B;color:#FD8E5B;font-size:14px">Gemma4-12B</th>

42

<th style="width:14.40%;padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #FD8E5B;color:#FD8E5B;font-size:14px">Gemma4-31B</th>

43

</tr></thead>

44

<tbody>

45

<tr><td colspan="6" style="padding:8px 12px;font-weight:600;color:#FD8E5B;border-bottom:1px solid rgba(253, 142, 91, 0.2);background:rgba(253, 142, 91, 0.1)">Agentic Coding</td></tr>

46

<tr>

47

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">Terminal-Bench 2.1 <sub><small>(Terminus-2)</small></sub></td>

48

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">43.1</td>

49

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">21.3</td>

50

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">41.4</td>

51

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">21</td>

52

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">42.1</td>

53

</tr>

54

<tr>

55

<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">Terminal-Bench 2.1 <sub><small>(Claude Code)</small></sub></td>

56

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">40.6</td>

57

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">18.9</td>

58

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">38.9</td>

59

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

60

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

61

</tr>

62

<tr>

63

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

64

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">69.4</td>

65

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">53.2</td>

66

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">70</td>

67

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">44.2</td>

68

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52</td>

69

</tr>

70

<tr>

71

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

72

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">42.9</td>

73

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">31.3</td>

74

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">44.6</td>

75

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">27.6</td>

76

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

77

</tr>

78

<tr>

79

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

80

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">52</td>

81

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">39.7</td>

82

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">60.3</td>

83

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">32.5</td>

84

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

85

</tr>

86

<tr>

87

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

88

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">27.2</td>

89

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">16.2</td>

90

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">20.5</td>

91

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">10.3</td>

92

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

93

</tr>

94

<tr>

95

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

96

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">63.1</td>

97

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">53.2</td>

98

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">65.4</td>

99

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">32.5</td>

100

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

101

</tr>

102

<tr>

103

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

104

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">17.9</td>

105

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">9.2</td>

106

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">13.2</td>

107

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

108

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

109

</tr>

110

<tr>

111

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

112

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">16.6</td>

113

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">4.3</td>

114

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">10.2</td>

115

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

116

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

117

</tr>

118

<tr>

119

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

120

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15);font-weight:600;color:#FD8E5B;background:rgba(253, 142, 91, 0.06)">15.3</td>

121

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">4.4</td>

122

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">9.8</td>

123

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

124

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

125

</tr>

126

</tbody>

127

</table>

128

129

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

130

* Terminal-Bench 2.1 (Terminus-2): We evaluate Terminal-Bench 2.1 using the Harbor/Terminus-2 framework with parser=json, temperature=1.0, top_p=1.0, and a 128K context window. Each run uses a 4-hour timeout with 32 CPU cores and 48GB RAM, and results are averaged over 5 runs. We adjust the Qwen chat template to ensure consistency between training and inference (https://huggingface.co/deepreinforce-ai/Ornith-1.0-397B/blob/main/chat_template.jinja), and modify Harbor to align with vLLM's reasoning_content key.<br/>

131

* Terminal-Bench 2.1 (Claude Code): We evaluate Terminal-Bench 2.1 using Claude Code 2.1.126 with parser=json, temperature=1.0, top_p=1.0, max_new_tokens=131072. Results are averaged over 5 runs. Again, Qwen chat template needs to be modified.<br/>

132

* SWE-Bench Verified, Pro and Multilingual: using OpenHands harness with temp=1.0, top_p=0.95, 256k context window.<br/>

133

* SWE Atlas QnA, RF, TW: using mini SWE agent harness with temp=1.0, top_p=0.95, 128K context window. Results are averaged over 5 runs.<br/>

134

* NL2Repo: with temperature=1.0, top_p=1.0, 400K context, 48K output and anti-hacking filters.<br/>

135

* ClawEval: An agentic code benchmark over real-user task distributions; temp=0.6 and 256K context.<br/>

136

</p>

137

138

</div>

139

140

## Quickstart

141

142

<div style="border-left:4px solid #FD8E5B;background:rgba(253,142,91,0.1);border-radius:6px;padding:12px 16px;font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;font-size:14px;line-height:1.6">

143

<div style="font-weight:700;color:#FD8E5B;margin-bottom:6px">📝 NOTE</div>

144

<p style="margin:0 0 10px"><b>Ornith-1.0-9B</b> is a <b>reasoning model</b>: by default the assistant turn opens with a <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">&lt;think&gt; … &lt;/think&gt;</code> block before the final answer. The serving recipes below enable a reasoning parser so the chain-of-thought is returned in a separate <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">reasoning_content</code> field, and a tool-call parser so the model's <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">&lt;tool_call&gt;</code> blocks are surfaced as OpenAI-style <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">tool_calls</code>.</p>

145

<p style="margin:0 0 6px">Serving Ornith-1.0-9B requires recent runtimes:</p>

146

<ul style="margin:0 0 10px;padding-left:20px">

147

<li><b>Transformers</b> ≥ 5.8.1</li>

148

<li><b>vLLM</b> ≥ 0.19.1</li>

149

<li><b>SGLang</b> ≥ 0.5.9</li>

150

</ul>

151

<p style="margin:0">Recommended sampling parameters: <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">temperature=0.6</code>, <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">top_p=0.95</code>, <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">top_k=20</code> (use <code style="background:rgba(253,142,91,0.15);padding:1px 5px;border-radius:4px">temperature=1.0</code> to reproduce the reported benchmark setup).</p>

152

</div>

153

154

155

### Serving Ornith-1.0-9B

156

157

Ornith-1.0-9B is a dense ~9B model (≈19 GB in bf16), so it serves comfortably on a **single 80GB GPU**. The recipes below stand up an OpenAI-compatible server; add `--tensor-parallel-size` / `--tp` if you want to shard across more GPUs.

158

159

#### vLLM

160

161

```bash

162

vllm serve deepreinforce-ai/Ornith-1.0-9B \

163

--served-model-name Ornith-1.0-9B \

164

--host 0.0.0.0 --port 8000 \

165

--max-model-len 262144 \

166

--gpu-memory-utilization 0.90 \

167

--enable-prefix-caching \

168

--enable-auto-tool-choice --tool-call-parser qwen3_xml \

169

--reasoning-parser qwen3 \

170

--trust-remote-code

171

```

172

173

#### SGLang

174

175

```bash

176

python -m sglang.launch_server \

177

--model-path deepreinforce-ai/Ornith-1.0-9B \

178

--served-model-name Ornith-1.0-9B \

179

--host 0.0.0.0 --port 8000 \

180

--context-length 262144 \

181

--mem-fraction-static 0.85 \

182

--tool-call-parser qwen3_coder \

183

--reasoning-parser qwen3

184

```

185

186

#### Hugging Face Transformers

187

188

For a quick local test (or to script offline generation), load the model directly with Transformers. Make sure you have a recent release installed — see the [Transformers installation guide](https://huggingface.co/docs/transformers/installation); Ornith-1.0-9B requires `transformers >= 5.8.1`.

189

190

```python

191

from transformers import AutoModelForCausalLM, AutoTokenizer

192

193

model_name = "deepreinforce-ai/Ornith-1.0-9B"

194

195

tokenizer = AutoTokenizer.from_pretrained(model_name)

196

model = AutoModelForCausalLM.from_pretrained(

197

model_name,

198

dtype="auto",

199

device_map="auto",

200

)

201

202

messages = [

203

{"role": "user", "content": "Write a Python function is_prime(n). Keep it short."}

204

]

205

text = tokenizer.apply_chat_template(

206

messages,

207

tokenize=False,

208

add_generation_prompt=True,

209

)

210

211

inputs = tokenizer(text, return_tensors="pt").to(model.device)

212

generated = model.generate(

213

**inputs,

214

max_new_tokens=512,

215

do_sample=True,

216

temperature=0.6,

217

top_p=0.95,

218

top_k=20,

219

)

220

output_ids = generated[0][inputs.input_ids.shape[1]:]

221

222

# The reply contains a <think> ... </think> reasoning block followed by the answer.

223

content = tokenizer.decode(output_ids, skip_special_tokens=True)

224

print(content)

225

```

226

227

To split the reasoning trace from the final answer, parse on the `</think>` marker:

228

229

```python

230

text = tokenizer.decode(output_ids, skip_special_tokens=True)

231

if "</think>" in text:

232

reasoning, answer = text.split("</think>", 1)

233

reasoning = reasoning.replace("<think>", "").strip()

234

answer = answer.strip()

235

else:

236

reasoning, answer = "", text.strip()

237

```

238

239

### Using Ornith-1.0-9B via the Chat Completions API

240

241

Once a vLLM or SGLang server is running, talk to it with any OpenAI-compatible client.

242

243

#### Basic Usage

244

245

```python

246

from openai import OpenAI

247

248

client = OpenAI(

249

base_url="http://localhost:8000/v1",

250

api_key="EMPTY", # any non-empty string works for a local server

251

)

252

253

response = client.chat.completions.create(

254

model="Ornith-1.0-9B",

255

messages=[

256

{"role": "user", "content": "Write a one-line Python lambda that squares a number."}

257

],

258

temperature=0.6,

259

top_p=0.95,

260

max_tokens=1024,

261

)

262

263

message = response.choices[0].message

264

# reasoning_content holds the <think> trace; content holds the final answer.

265

print("reasoning:", getattr(message, "reasoning_content", None))

266

print("answer:", message.content)

267

```

268

269

You can also stream tokens, or hand the model tools — Ornith-1.0-9B emits well-formed function calls that the server parses into the standard `tool_calls` field:

270

271

```python

272

tools = [

273

{

274

"type": "function",

275

"function": {

276

"name": "get_weather",

277

"description": "Get the current weather for a city",

278

"parameters": {

279

"type": "object",

280

"properties": {"city": {"type": "string"}},

281

"required": ["city"],

282

},

283

},

284

}

285

]

286

287

response = client.chat.completions.create(

288

model="Ornith-1.0-9B",

289

messages=[{"role": "user", "content": "What is the weather in Paris right now?"}],

290

tools=tools,

291

tool_choice="auto",

292

temperature=0.6,

293

max_tokens=2048,

294

)

295

296

tool_call = response.choices[0].message.tool_calls[0]

297

print(tool_call.function.name, tool_call.function.arguments)

298

# -> get_weather {"city": "Paris"}

299

```

300

301

You can point any OpenAI-compatible SDK (Python, Node.js, etc.) or `curl` at the same `/v1/chat/completions` endpoint.

302

303

304

## Agentic Usage

305

306

Ornith-1.0-9B excels in tool-calling and agentic coding capabilities.

307

308

### Agent Frameworks

309

310

Because Ornith-1.0-9B exposes an OpenAI-compatible endpoint with tool calling, it works out of the box with standard agent frameworks. Below is a minimal example that connects Ornith-1.0-9B to tools through an MCP server.

311

312

```python

313

import os

314

from openai import OpenAI

315

316

client = OpenAI(

317

base_url=os.getenv("OPENAI_BASE_URL", "http://localhost:8000/v1"),

318

api_key=os.getenv("OPENAI_API_KEY", "EMPTY"),

319

)

320

321

tools = [

322

{

323

"type": "function",

324

"function": {

325

"name": "run_shell",

326

"description": "Run a shell command and return its output.",

327

"parameters": {

328

"type": "object",

329

"properties": {

330

"command": {"type": "string", "description": "The command to run"}

331

},

332

"required": ["command"],

333

},

334

},

335

}

336

]

337

338

messages = [{"role": "user", "content": "List the Python files in the current directory."}]

339

340

response = client.chat.completions.create(

341

model="deepreinforce-ai/Ornith-1.0-9B",

342

messages=messages,

343

tools=tools,

344

temperature=0.6,

345

top_p=0.95,

346

)

347

print(response.choices[0].message)

348

```

349

350

**Examples of using Ornith with agent harness:**

351

352

#### Hermes Agent

353

```bash

354

# Hermes talks to any OpenAI-compatible endpoint — point it at your Ornith server.

355

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

356

export OPENAI_API_KEY="EMPTY"

357

export MODEL="deepreinforce-ai/Ornith-1.0-9B"

358

```

359

360

361

#### Atomic.chat / Ollama / llama.cpp

362

```bash

363

# Both runtimes load a GGUF build of Ornith (publish one at deepreinforce-ai/Ornith-1.0-9B-GGUF).

364

365

# llama.cpp — serve an OpenAI-compatible API on port 8000.

366

llama-server -hf deepreinforce-ai/Ornith-1.0-9B-GGUF --port 8000 -c 262144

367

368

# Ollama — pull and chat with the same GGUF straight from Hugging Face.

369

ollama run hf.co/deepreinforce-ai/Ornith-1.0-9B-GGUF

370

```

371

372

#### OpenClaw

373

374

```bash

375

# OpenClaw talks to any OpenAI-compatible endpoint — point it at your Ornith server.

376

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

377

export OPENAI_API_KEY="EMPTY"

378

export OPENAI_MODEL="deepreinforce-ai/Ornith-1.0-9B"

379

```

380

381

#### Unsloth Studio

382

383

```bash

384

pip install unsloth

385

386

# Load Ornith for fast local inference or fine-tuning (Python):

387

# from unsloth import FastLanguageModel

388

# model, tokenizer = FastLanguageModel.from_pretrained(

389

# "deepreinforce-ai/Ornith-1.0-9B",

390

# max_seq_length=262144,

391

# load_in_4bit=True,

392

# )

393

```

394

395

#### OpenHands

396

```bash

397

pip install openhands-ai

398

399

# OpenHands routes through LiteLLM; the "openai/" prefix selects the OpenAI-compatible path.

400

export LLM_MODEL="openai/deepreinforce-ai/Ornith-1.0-9B"

401

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

402

export LLM_API_KEY="EMPTY"

403

404

# Launch the CLI (or run the official OpenHands Docker image with the same env vars).

405

openhands

406

```

407

408

### Coding CLIs

409

410

Ornith-1.0-9B is optimized for terminal-based coding agents. Point any OpenAI-compatible coding CLI at your Ornith-1.0-9B endpoint (set `OPENAI_BASE_URL` and `OPENAI_API_KEY`) to understand large codebases, automate tedious work, and ship faster.

411

412

#### OpenCode

413

```bash

414

# Register your local Ornith endpoint as a provider in ~/.config/opencode/opencode.json:

415

#

416

# {

417

# "$schema": "https://opencode.ai/config.json",

418

# "provider": {

419

# "ornith": {

420

# "npm": "@ai-sdk/openai-compatible",

421

# "name": "Ornith (local)",

422

# "options": { "baseURL": "http://localhost:8000/v1", "apiKey": "EMPTY" },

423

# "models": { "deepreinforce-ai/Ornith-1.0-9B": { "name": "Ornith-1.0-9B" } }

424

# }

425

# }

426

# }

427

428

opencode

429

```

430

431

### Citation

432

433

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

434

435

```bibtex

436

@misc{ornith_9b,

437

title = {{Ornith-1.0-9B}: Agentic Coding, Open to All},

438

url = {https://deep-reinforce.com/ornith_1_0.html},

439

author = {{DeepReinforce Team}},

440

year = {2026}

441

}

442

```