README.md · Ornith-1.0-35B

1

---

2

library_name: transformers

3

license: mit

4

license_link: https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B/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

# Ornith-1.0-35B

14

15

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

16

17

Highlights:

18

19

- **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.

20

- **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.

21

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

22

23

<img style="width: 100%; max-width: 900px;" src="assets/ornith_35b_eval.png" alt="Ornith 35B Benchmark Results" title="Ornith 35B Benchmark Results">

24

25

## Ornith 1.0 35B

26

27

This model card documents **Ornith-1.0-35B**, the lightweight member of the Ornith family, designed for efficient single-GPU deployment.

28

29

30

### Benchmarks

31

32

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

33

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

34

<thead><tr>

35

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

36

<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-1.0-35B</th>

37

<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>

38

<th style="width:14.40%;padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #FD8E5B;color:#FD8E5B;font-size:14px">Qwen3.6-35B</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">Gemma4-31B</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">Qwen3.5-397B</th>

41

</tr></thead>

42

<tbody>

43

<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>

44

<tr>

45

<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>

46

<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)">64.2</td>

47

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

48

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

49

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

50

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

51

</tr>

52

<tr>

53

<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>

54

<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)">62.8</td>

55

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

56

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">49.2</td>

57

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

58

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">48.6</td>

59

</tr>

60

<tr>

61

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

62

<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)">75.6</td>

63

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

64

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

65

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

66

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">76.4</td>

67

</tr>

68

<tr>

69

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

70

<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)">50.4</td>

71

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

72

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

73

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

74

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">51.6</td>

75

</tr>

76

<tr>

77

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

78

<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)">69.3</td>

79

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

80

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

81

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

82

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

83

</tr>

84

<tr>

85

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

86

<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)">34.6</td>

87

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

88

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

89

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

90

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">36.8</td>

91

</tr>

92

<tr>

93

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

94

<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)">69.8</td>

95

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

96

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

97

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

98

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

99

</tr>

100

<tr>

101

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

102

<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)">37.1</td>

103

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

104

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

105

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

106

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">20.4</td>

107

</tr>

108

<tr>

109

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

110

<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)">29.7</td>

111

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

112

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">11.4</td>

113

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

114

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">18.4</td>

115

</tr>

116

<tr>

117

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

118

<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)">27.8</td>

119

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

120

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">13.3</td>

121

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

122

<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">18.5</td>

123

</tr>

124

</tbody>

125

</table>

126

127

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

128

* 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/>

129

* 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/>

130

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

131

* 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/>

132

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

133

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

134

</p>

135

136

</div>

137

138

139

140

141

## Quickstart

142

143

144

<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">

145

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

146

<p style="margin:0 0 10px"><b>Ornith-1.0-35B</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>

147

<p style="margin:0 0 6px">Serving Ornith-1.0-35B requires recent runtimes:</p>

148

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

149

<li><b>Transformers</b> ≥ 5.8.1</li>

150

<li><b>vLLM</b> ≥ 0.19.1</li>

151

<li><b>SGLang</b> ≥ 0.5.9</li>

152

</ul>

153

</div>

154

155

### Serving Ornith-1.0-35B

156

157

The two recipes below stand up an OpenAI-compatible server on a single 8×80GB GPU node (tensor-parallel 8). Adjust `--tensor-parallel-size` / `--tp` to the number of GPUs you have.

158

159

#### vLLM

160

161

```bash

162

vllm serve deepreinforce-ai/Ornith-1.0-35B \

163

--served-model-name Ornith-1.0-35B \

164

--tensor-parallel-size 8 \

165

--host 0.0.0.0 --port 8000 \

166

--max-model-len 262144 \

167

--gpu-memory-utilization 0.90 \

168

--enable-prefix-caching \

169

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

170

--reasoning-parser qwen3 \

171

--trust-remote-code

172

```

173

174

#### SGLang

175

176

```bash

177

python -m sglang.launch_server \

178

--model-path deepreinforce-ai/Ornith-1.0-35B \

179

--served-model-name Ornith-1.0-35B \

180

--tp 8 \

181

--host 0.0.0.0 --port 8000 \

182

--context-length 262144 \

183

--mem-fraction-static 0.85 \

184

--tool-call-parser qwen3_coder \

185

--reasoning-parser qwen3

186

```

187

188

#### Hugging Face Transformers

189

190

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-1.0-35B requires `transformers >= 5.8.1`.

191

192

```python

193

from transformers import AutoModelForCausalLM, AutoTokenizer

194

195

model_name = "deepreinforce-ai/Ornith-1.0-35B"

196

197

tokenizer = AutoTokenizer.from_pretrained(model_name)

198

model = AutoModelForCausalLM.from_pretrained(

199

model_name,

200

dtype="auto",

201

device_map="auto",

202

)

203

204

messages = [

205

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

206

]

207

text = tokenizer.apply_chat_template(

208

messages,

209

tokenize=False,

210

add_generation_prompt=True,

211

)

212

213

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

214

generated = model.generate(

215

**inputs,

216

max_new_tokens=512,

217

do_sample=True,

218

temperature=0.6,

219

top_p=0.95,

220

top_k=20,

221

)

222

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

223

224

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

225

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

226

print(content)

227

```

228

229

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

230

231

```python

232

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

233

if "</think>" in text:

234

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

235

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

236

answer = answer.strip()

237

else:

238

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

239

```

240

241

### Using Ornith-1.0-35B via the Chat Completions API

242

243

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

244

245

#### Basic Usage

246

247

```python

248

from openai import OpenAI

249

250

client = OpenAI(

251

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

252

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

253

)

254

255

response = client.chat.completions.create(

256

model="Ornith-1.0-35B",

257

messages=[

258

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

259

],

260

temperature=0.6,

261

top_p=0.95,

262

max_tokens=1024,

263

)

264

265

message = response.choices[0].message

266

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

267

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

268

print("answer:", message.content)

269

```

270

271

You can also stream tokens, or hand the model tools — Ornith-1.0-35B emits well-formed function calls that the server parses into the standard `tool_calls` field:

272

273

```python

274

tools = [

275

{

276

"type": "function",

277

"function": {

278

"name": "get_weather",

279

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

280

"parameters": {

281

"type": "object",

282

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

283

"required": ["city"],

284

},

285

},

286

}

287

]

288

289

response = client.chat.completions.create(

290

model="Ornith-1.0-35B",

291

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

292

tools=tools,

293

tool_choice="auto",

294

temperature=0.6,

295

max_tokens=2048,

296

)

297

298

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

299

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

300

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

301

```

302

303

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

304

305

## Agentic Usage

306

307

Ornith-1.0-35B excels in tool-calling and agentic coding capabilities.

308

309

### Agent Frameworks

310

311

Because Ornith-1.0-35B 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-1.0-35B to tools through an MCP server.

312

313

```python

314

import os

315

from openai import OpenAI

316

317

client = OpenAI(

318

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

319

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

320

)

321

322

tools = [

323

{

324

"type": "function",

325

"function": {

326

"name": "run_shell",

327

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

328

"parameters": {

329

"type": "object",

330

"properties": {

331

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

332

},

333

"required": ["command"],

334

},

335

},

336

}

337

]

338

339

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

340

341

response = client.chat.completions.create(

342

model="deepreinforce-ai/Ornith-1.0-35B",

343

messages=messages,

344

tools=tools,

345

temperature=0.6,

346

top_p=0.95,

347

)

348

print(response.choices[0].message)

349

```

350

351

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

352

353

#### Hermes Agent

354

```bash

355

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

356

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

357

export OPENAI_API_KEY="EMPTY"

358

export MODEL="deepreinforce-ai/Ornith-1.0-35B"

359

```

360

361

362

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

363

```bash

364

# Both runtimes load a GGUF build of Ornith (publish one at deepreinforce-ai/Ornith-1.0-35B-GGUF).

365

366

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

367

llama-server -hf deepreinforce-ai/Ornith-1.0-35B-GGUF --port 8000 -c 262144

368

369

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

370

ollama run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF

371

```

372

373

#### OpenClaw

374

375

```bash

376

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

377

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

378

export OPENAI_API_KEY="EMPTY"

379

export OPENAI_MODEL="deepreinforce-ai/Ornith-1.0-35B"

380

```

381

382

#### Unsloth Studio

383

384

```bash

385

pip install unsloth

386

387

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

388

# from unsloth import FastLanguageModel

389

# model, tokenizer = FastLanguageModel.from_pretrained(

390

# "deepreinforce-ai/Ornith-1.0-35B",

391

# max_seq_length=262144,

392

# load_in_4bit=True,

393

# )

394

```

395

396

#### OpenHands

397

```bash

398

pip install openhands-ai

399

400

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

401

export LLM_MODEL="openai/deepreinforce-ai/Ornith-1.0-35B"

402

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

403

export LLM_API_KEY="EMPTY"

404

405

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

406

openhands

407

```

408

409

### Coding CLIs

410

411

Ornith-1.0-35B is optimized for terminal-based coding agents. Point any OpenAI-compatible coding CLI at your Ornith-1.0-35B endpoint (set `OPENAI_BASE_URL` and `OPENAI_API_KEY`) to understand large codebases, automate tedious work, and ship faster.

412

413

#### OpenCode

414

```bash

415

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

416

#

417

# {

418

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

419

# "provider": {

420

# "ornith": {

421

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

422

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

423

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

424

# "models": { "deepreinforce-ai/Ornith-1.0-35B": { "name": "Ornith-1.0-35B" } }

425

# }

426

# }

427

# }

428

429

opencode

430

```

431

432

433

### Citation

434

435

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

436

437

```bibtex

438

@misc{ornith-35b,

439

title = {{Ornith-1.0-35B}: Agentic Coding, Open to All},

440

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

441

author = {{DeepReinforce Team}},

442

year = {2026}

443

}

444

```

445