Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,505 @@
|
|
| 1 |
---
|
|
|
|
| 2 |
license: mit
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
library_name: transformers
|
| 3 |
license: mit
|
| 4 |
+
license_link: https://huggingface.co/deepreinforce-ai/Ornith-1.5-9B/blob/main/LICENSE
|
| 5 |
+
pipeline_tag: text-generation
|
| 6 |
---
|
| 7 |
+
|
| 8 |
+
|
| 9 |
+
<img width="600px" src="assets/ornith_logo.png">
|
| 10 |
+
|
| 11 |
+
[](https://deep-reinforce.com/ornith.html)
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
# Ornith-1.5-9B
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
Chirp Chirp! 🐦 We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement.
|
| 18 |
+
|
| 19 |
+
Highlights:
|
| 20 |
+
|
| 21 |
+
- **End-to-End Self-Improvement**: Ornith-1.5 extends the self-scaffolding framework introduced in Ornith-1.0 into a complete self-improvement loop: the model proposes new tasks, generates task-specific scaffolds, and produces solution rollouts for reinforcement learning, continuously creating new learning experiences from which it can improve. Task generation, scaffold construction, and solution rollouts are all optimized jointly with GRPO, so the model continually expands its own curriculum and adapts its problem-solving strategies.
|
| 22 |
+
- **State-of-the-Art General Intelligence**: Designed for strong general-purpose intelligence across reasoning, agentic, and coding tasks, Ornith-1.5 achieves state-of-the-art performance among open-source models of comparable size across a broad range of benchmarks.
|
| 23 |
+
- **Edge-Deployable 9B**: Ornith-1.5-9B achieves 47.0 on Terminal-Bench 2.1 (Claude Code) and 70.6 on SWE-Bench Verified — matching or exceeding much larger models such as Gemma 4-31B and Qwen 3.6-35B. With its quantized **Ornith-1.5-9B-Mobile** version, it can be readily deployed on iPhone and Android devices.
|
| 24 |
+
- **Licence**: MIT licensed, globally accessible, and free from regional limitations.
|
| 25 |
+
|
| 26 |
+
<img style="width: 100%; max-width: 900px;" src="assets/ornith_9b_eval.png" alt="Ornith 1.5 9B Benchmark Results" title="Ornith 1.5 9B Benchmark Results">
|
| 27 |
+
|
| 28 |
+
## Ornith 1.5 9B
|
| 29 |
+
|
| 30 |
+
This model card documents **Ornith-1.5-9B**, the most lightweight member of the Ornith-1.5 family — a compact 9B dense model designed for efficient single-GPU deployment, and edge-deployable on mobile devices via its quantized Ornith-1.5-9B-Mobile variant.
|
| 31 |
+
|
| 32 |
+
### Benchmarks
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;width:100%;margin:0 auto;padding:16px 0">
|
| 36 |
+
<table style="width:100%;table-layout:fixed;border-collapse:collapse;font-size:13px">
|
| 37 |
+
<thead><tr>
|
| 38 |
+
<th style="width:28%;padding:10px 7px;text-align:left;font-weight:600;border-bottom:2px solid #FD8E5B;color:#FD8E5B"></th>
|
| 39 |
+
<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-1.5-9B</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">Ornith-1.0-9B</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">Qwen3.5-9B</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">Qwen3.6-35B-A3B</th>
|
| 43 |
+
<th style="width:14.40%;padding:10px 7px;text-align:center;font-weight:500;border-bottom:2px solid #FD8E5B;color:#FD8E5B;font-size:14px">Gemma-4-31B <sub><small>(dense)</small></sub></th>
|
| 44 |
+
</tr></thead>
|
| 45 |
+
<tbody>
|
| 46 |
+
<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)">Coding</td></tr>
|
| 47 |
+
<tr>
|
| 48 |
+
<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>
|
| 49 |
+
<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)">46.2</td>
|
| 50 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">43.1</td>
|
| 51 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">21.3</td>
|
| 52 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.5</td>
|
| 53 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">42.1</td>
|
| 54 |
+
</tr>
|
| 55 |
+
<tr>
|
| 56 |
+
<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>
|
| 57 |
+
<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)">47</td>
|
| 58 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">40.6</td>
|
| 59 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">18.9</td>
|
| 60 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">49.2</td>
|
| 61 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">-</td>
|
| 62 |
+
</tr>
|
| 63 |
+
<tr>
|
| 64 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SWE-bench Verified</td>
|
| 65 |
+
<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)">70.6</td>
|
| 66 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">69.4</td>
|
| 67 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">53.2</td>
|
| 68 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">73.4</td>
|
| 69 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52</td>
|
| 70 |
+
</tr>
|
| 71 |
+
<tr>
|
| 72 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SWE-bench Pro</td>
|
| 73 |
+
<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)">47.5</td>
|
| 74 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">42.9</td>
|
| 75 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">31.3</td>
|
| 76 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">49.5</td>
|
| 77 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">35.7</td>
|
| 78 |
+
</tr>
|
| 79 |
+
<tr>
|
| 80 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SWE-bench Multilingual</td>
|
| 81 |
+
<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)">54.4</td>
|
| 82 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52</td>
|
| 83 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">39.7</td>
|
| 84 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">67.2</td>
|
| 85 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">51.7</td>
|
| 86 |
+
</tr>
|
| 87 |
+
<tr>
|
| 88 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">DeepSWE</td>
|
| 89 |
+
<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)">8</td>
|
| 90 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">0</td>
|
| 91 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">0</td>
|
| 92 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">0</td>
|
| 93 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">-</td>
|
| 94 |
+
</tr>
|
| 95 |
+
<tr>
|
| 96 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">NL2Repo</td>
|
| 97 |
+
<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)">32.4</td>
|
| 98 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">27.2</td>
|
| 99 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">16.2</td>
|
| 100 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">29.4</td>
|
| 101 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">15.5</td>
|
| 102 |
+
</tr>
|
| 103 |
+
<tr>
|
| 104 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">SWE Atlas - QnA</td>
|
| 105 |
+
<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)">20.6</td>
|
| 106 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">17.9</td>
|
| 107 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">9.2</td>
|
| 108 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">15.5</td>
|
| 109 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">-</td>
|
| 110 |
+
</tr>
|
| 111 |
+
<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)">Reasoning</td></tr>
|
| 112 |
+
<tr>
|
| 113 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">HLE <sub><small>(no tools)</small></sub></td>
|
| 114 |
+
<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)">20.2</td>
|
| 115 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">16.8</td>
|
| 116 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">14.7</td>
|
| 117 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">21.4</td>
|
| 118 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">19.5</td>
|
| 119 |
+
</tr>
|
| 120 |
+
<tr>
|
| 121 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">HLE <sub><small>(with tools)</small></sub></td>
|
| 122 |
+
<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)">30.5</td>
|
| 123 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">26.4</td>
|
| 124 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">24.5</td>
|
| 125 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">28.9</td>
|
| 126 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">26.5</td>
|
| 127 |
+
</tr>
|
| 128 |
+
<tr>
|
| 129 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">GPQA Diamond</td>
|
| 130 |
+
<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)">86.4</td>
|
| 131 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">82.5</td>
|
| 132 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">81.7</td>
|
| 133 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">86</td>
|
| 134 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">84.3</td>
|
| 135 |
+
</tr>
|
| 136 |
+
<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</td></tr>
|
| 137 |
+
<tr>
|
| 138 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">MCP-Atlas</td>
|
| 139 |
+
<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)">54.2</td>
|
| 140 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">49.4</td>
|
| 141 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">46.8</td>
|
| 142 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">62.8</td>
|
| 143 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">55</td>
|
| 144 |
+
</tr>
|
| 145 |
+
<tr>
|
| 146 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">Toolathlon-Verified</td>
|
| 147 |
+
<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)">41.2</td>
|
| 148 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">33.4</td>
|
| 149 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">29.6</td>
|
| 150 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">41.7</td>
|
| 151 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">52.8</td>
|
| 152 |
+
</tr>
|
| 153 |
+
<tr>
|
| 154 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">WideSearch</td>
|
| 155 |
+
<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)">59.5</td>
|
| 156 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">55.8</td>
|
| 157 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">53.6</td>
|
| 158 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">60.1</td>
|
| 159 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">54.2</td>
|
| 160 |
+
</tr>
|
| 161 |
+
<tr>
|
| 162 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">BrowseComp</td>
|
| 163 |
+
<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)">56.4</td>
|
| 164 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">44.8</td>
|
| 165 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">41.5</td>
|
| 166 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">62</td>
|
| 167 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">-</td>
|
| 168 |
+
</tr>
|
| 169 |
+
<tr>
|
| 170 |
+
<td style="padding:7px 7px;padding-left:20px;border-bottom:1px solid rgba(128, 128, 128, 0.15);">ClawEval</td>
|
| 171 |
+
<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)">66.5</td>
|
| 172 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">63.1</td>
|
| 173 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">53.2</td>
|
| 174 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">68.7</td>
|
| 175 |
+
<td style="padding:7px 7px;text-align:center;border-bottom:1px solid rgba(128, 128, 128, 0.15)">48.5</td>
|
| 176 |
+
</tr>
|
| 177 |
+
</tbody>
|
| 178 |
+
</table>
|
| 179 |
+
|
| 180 |
+
<p style="margin-top:12px;font-size:10px;opacity:0.7">
|
| 181 |
+
* All results reported for Ornith-1.5 are averaged over five independent runs.<br/>
|
| 182 |
+
* 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-1.5-9B/blob/main/chat_template.jinja), and modify Harbor to align with vLLM's reasoning_content key.<br/>
|
| 183 |
+
* 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/>
|
| 184 |
+
* SWE-Bench Verified, Pro and Multilingual: using OpenHands harness with temp=1.0, top_p=0.95, 256k context window. Anti-hacking safeguards are applied throughout evaluation: Git history is removed from the local repository image to prevent access to prior solutions or commits; network access is disabled, preventing the model from retrieving external information or resources.<br/>
|
| 185 |
+
* DeepSWE: Evaluated using the Claude Code harness with temperature=1.0, top_p=0.95, and a 256K context window.<br/>
|
| 186 |
+
* SWE Atlas QnA: using mini SWE agent harness with temp=1.0, top_p=0.95, 128K context window. Results are averaged over 5 runs.<br/>
|
| 187 |
+
* NL2Repo: with temperature=1.0, top_p=1.0, 400K context, 48K output. Access to specified GitHub repositories and pip packages is blocked to prevent reward hacking.<br/>
|
| 188 |
+
* HLE: Evaluated using Claude 4.6 Opus as the judge model.<br/>
|
| 189 |
+
* MCP-Atlas: All models were evaluated in thinking mode on the 500-task public subset, with a 10-minute timeout per task. We use Claude 4.8 Opus as the judge model.<br/>
|
| 190 |
+
* Toolathlon-Verified: We use the official evaluation service with the maximum token limit set to 128K.<br/>
|
| 191 |
+
* ClawEval: An agentic code benchmark over real-user task distributions; temp=0.6 and 256K context.<br/>
|
| 192 |
+
</p>
|
| 193 |
+
|
| 194 |
+
</div>
|
| 195 |
+
|
| 196 |
+
## Self-Improvement through Self-Generated Tasks, Harnesses, and Solutions
|
| 197 |
+
|
| 198 |
+
Ornith-1.5 extends Ornith-1.0 by expanding the self-improvement loop from scaffold and rollout optimization to jointly optimizing task generation, scaffold construction, and solution rollouts. Rather than relying on a fixed set of human-curated tasks and manually designed harnesses, Ornith-1.5 continuously generates new training tasks, discovers effective strategies for solving them, and improves the policy through reinforcement learning.
|
| 199 |
+
|
| 200 |
+
Each training cycle proceeds in three stages. Given an environment or codebase, high-level instructions about the task type, and access to the model's previous task-solving history, the system proposes progressively harder tasks that go beyond what the model has already solved. For each task, the model then generates or refines a task-specific scaffold — the instructions, tools, decomposition strategy, and orchestration used to approach the problem. Conditioned on the task and scaffold, the policy produces a solution rollout. Reward from the rollout is propagated across all three stages, so the system learns not only to produce better solutions, but also to generate more useful training tasks and construct more effective scaffolds.
|
| 201 |
+
|
| 202 |
+
Repeated over training, this creates a closed self-improvement loop in which stronger policies enable the generation of harder and more informative tasks, evolving scaffolds discover better ways to elicit the model's capabilities, and higher-quality rollouts provide increasingly effective learning signals. For more details on the task, harness, and rollout reward design, please refer to our [blog](https://deep-reinforce.com/ornith.html).
|
| 203 |
+
|
| 204 |
+
## Quickstart
|
| 205 |
+
|
| 206 |
+
<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">
|
| 207 |
+
<div style="font-weight:700;color:#FD8E5B;margin-bottom:6px">📝 NOTE</div>
|
| 208 |
+
<p style="margin:0 0 10px"><b>Ornith-1.5-9B</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"><think> … </think></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"><tool_call></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>
|
| 209 |
+
<p style="margin:0 0 6px">Serving Ornith-1.5-9B requires recent runtimes:</p>
|
| 210 |
+
<ul style="margin:0 0 10px;padding-left:20px">
|
| 211 |
+
<li><b>Transformers</b> ≥ 5.8.1</li>
|
| 212 |
+
<li><b>vLLM</b> ≥ 0.19.1</li>
|
| 213 |
+
<li><b>SGLang</b> ≥ 0.5.9</li>
|
| 214 |
+
</ul>
|
| 215 |
+
<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>
|
| 216 |
+
</div>
|
| 217 |
+
|
| 218 |
+
|
| 219 |
+
### Serving Ornith-1.5-9B
|
| 220 |
+
|
| 221 |
+
Ornith-1.5-9B 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.
|
| 222 |
+
|
| 223 |
+
#### vLLM
|
| 224 |
+
|
| 225 |
+
```bash
|
| 226 |
+
vllm serve deepreinforce-ai/Ornith-1.5-9B \
|
| 227 |
+
--served-model-name Ornith-1.5-9B \
|
| 228 |
+
--host 0.0.0.0 --port 8000 \
|
| 229 |
+
--max-model-len 262144 \
|
| 230 |
+
--gpu-memory-utilization 0.90 \
|
| 231 |
+
--enable-prefix-caching \
|
| 232 |
+
--enable-auto-tool-choice --tool-call-parser qwen3_xml \
|
| 233 |
+
--reasoning-parser qwen3 \
|
| 234 |
+
--trust-remote-code
|
| 235 |
+
```
|
| 236 |
+
|
| 237 |
+
#### SGLang
|
| 238 |
+
|
| 239 |
+
```bash
|
| 240 |
+
python -m sglang.launch_server \
|
| 241 |
+
--model-path deepreinforce-ai/Ornith-1.5-9B \
|
| 242 |
+
--served-model-name Ornith-1.5-9B \
|
| 243 |
+
--host 0.0.0.0 --port 8000 \
|
| 244 |
+
--context-length 262144 \
|
| 245 |
+
--mem-fraction-static 0.85 \
|
| 246 |
+
--tool-call-parser qwen3_coder \
|
| 247 |
+
--reasoning-parser qwen3
|
| 248 |
+
```
|
| 249 |
+
|
| 250 |
+
#### Hugging Face Transformers
|
| 251 |
+
|
| 252 |
+
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-1.5-9B requires `transformers >= 5.8.1`.
|
| 253 |
+
|
| 254 |
+
```python
|
| 255 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 256 |
+
|
| 257 |
+
model_name = "deepreinforce-ai/Ornith-1.5-9B"
|
| 258 |
+
|
| 259 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 260 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 261 |
+
model_name,
|
| 262 |
+
dtype="auto",
|
| 263 |
+
device_map="auto",
|
| 264 |
+
)
|
| 265 |
+
|
| 266 |
+
messages = [
|
| 267 |
+
{"role": "user", "content": "Write a Python function is_prime(n). Keep it short."}
|
| 268 |
+
]
|
| 269 |
+
text = tokenizer.apply_chat_template(
|
| 270 |
+
messages,
|
| 271 |
+
tokenize=False,
|
| 272 |
+
add_generation_prompt=True,
|
| 273 |
+
)
|
| 274 |
+
|
| 275 |
+
inputs = tokenizer(text, return_tensors="pt").to(model.device)
|
| 276 |
+
generated = model.generate(
|
| 277 |
+
**inputs,
|
| 278 |
+
max_new_tokens=512,
|
| 279 |
+
do_sample=True,
|
| 280 |
+
temperature=0.6,
|
| 281 |
+
top_p=0.95,
|
| 282 |
+
top_k=20,
|
| 283 |
+
)
|
| 284 |
+
output_ids = generated[0][inputs.input_ids.shape[1]:]
|
| 285 |
+
|
| 286 |
+
# The reply contains a <think> ... </think> reasoning block followed by the answer.
|
| 287 |
+
content = tokenizer.decode(output_ids, skip_special_tokens=True)
|
| 288 |
+
print(content)
|
| 289 |
+
```
|
| 290 |
+
|
| 291 |
+
To split the reasoning trace from the final answer, parse on the `</think>` marker:
|
| 292 |
+
|
| 293 |
+
```python
|
| 294 |
+
text = tokenizer.decode(output_ids, skip_special_tokens=True)
|
| 295 |
+
if "</think>" in text:
|
| 296 |
+
reasoning, answer = text.split("</think>", 1)
|
| 297 |
+
reasoning = reasoning.replace("<think>", "").strip()
|
| 298 |
+
answer = answer.strip()
|
| 299 |
+
else:
|
| 300 |
+
reasoning, answer = "", text.strip()
|
| 301 |
+
```
|
| 302 |
+
|
| 303 |
+
### Using Ornith-1.5-9B via the Chat Completions API
|
| 304 |
+
|
| 305 |
+
Once a vLLM or SGLang server is running, talk to it with any OpenAI-compatible client.
|
| 306 |
+
|
| 307 |
+
#### Basic Usage
|
| 308 |
+
|
| 309 |
+
```python
|
| 310 |
+
from openai import OpenAI
|
| 311 |
+
|
| 312 |
+
client = OpenAI(
|
| 313 |
+
base_url="http://localhost:8000/v1",
|
| 314 |
+
api_key="EMPTY", # any non-empty string works for a local server
|
| 315 |
+
)
|
| 316 |
+
|
| 317 |
+
response = client.chat.completions.create(
|
| 318 |
+
model="Ornith-1.5-9B",
|
| 319 |
+
messages=[
|
| 320 |
+
{"role": "user", "content": "Write a one-line Python lambda that squares a number."}
|
| 321 |
+
],
|
| 322 |
+
temperature=0.6,
|
| 323 |
+
top_p=0.95,
|
| 324 |
+
max_tokens=1024,
|
| 325 |
+
)
|
| 326 |
+
|
| 327 |
+
message = response.choices[0].message
|
| 328 |
+
# reasoning_content holds the <think> trace; content holds the final answer.
|
| 329 |
+
print("reasoning:", getattr(message, "reasoning_content", None))
|
| 330 |
+
print("answer:", message.content)
|
| 331 |
+
```
|
| 332 |
+
|
| 333 |
+
You can also stream tokens, or hand the model tools — Ornith-1.5-9B emits well-formed function calls that the server parses into the standard `tool_calls` field:
|
| 334 |
+
|
| 335 |
+
```python
|
| 336 |
+
tools = [
|
| 337 |
+
{
|
| 338 |
+
"type": "function",
|
| 339 |
+
"function": {
|
| 340 |
+
"name": "get_weather",
|
| 341 |
+
"description": "Get the current weather for a city",
|
| 342 |
+
"parameters": {
|
| 343 |
+
"type": "object",
|
| 344 |
+
"properties": {"city": {"type": "string"}},
|
| 345 |
+
"required": ["city"],
|
| 346 |
+
},
|
| 347 |
+
},
|
| 348 |
+
}
|
| 349 |
+
]
|
| 350 |
+
|
| 351 |
+
response = client.chat.completions.create(
|
| 352 |
+
model="Ornith-1.5-9B",
|
| 353 |
+
messages=[{"role": "user", "content": "What is the weather in Paris right now?"}],
|
| 354 |
+
tools=tools,
|
| 355 |
+
tool_choice="auto",
|
| 356 |
+
temperature=0.6,
|
| 357 |
+
max_tokens=2048,
|
| 358 |
+
)
|
| 359 |
+
|
| 360 |
+
tool_call = response.choices[0].message.tool_calls[0]
|
| 361 |
+
print(tool_call.function.name, tool_call.function.arguments)
|
| 362 |
+
# -> get_weather {"city": "Paris"}
|
| 363 |
+
```
|
| 364 |
+
|
| 365 |
+
You can point any OpenAI-compatible SDK (Python, Node.js, etc.) or `curl` at the same `/v1/chat/completions` endpoint.
|
| 366 |
+
|
| 367 |
+
## Agentic Usage
|
| 368 |
+
|
| 369 |
+
Ornith-1.5-9B excels in tool-calling and agentic coding capabilities.
|
| 370 |
+
|
| 371 |
+
### Agent Frameworks
|
| 372 |
+
|
| 373 |
+
Because Ornith-1.5-9B 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-1.5-9B to tools through an MCP server.
|
| 374 |
+
|
| 375 |
+
```python
|
| 376 |
+
import os
|
| 377 |
+
from openai import OpenAI
|
| 378 |
+
|
| 379 |
+
client = OpenAI(
|
| 380 |
+
base_url=os.getenv("OPENAI_BASE_URL", "http://localhost:8000/v1"),
|
| 381 |
+
api_key=os.getenv("OPENAI_API_KEY", "EMPTY"),
|
| 382 |
+
)
|
| 383 |
+
|
| 384 |
+
tools = [
|
| 385 |
+
{
|
| 386 |
+
"type": "function",
|
| 387 |
+
"function": {
|
| 388 |
+
"name": "run_shell",
|
| 389 |
+
"description": "Run a shell command and return its output.",
|
| 390 |
+
"parameters": {
|
| 391 |
+
"type": "object",
|
| 392 |
+
"properties": {
|
| 393 |
+
"command": {"type": "string", "description": "The command to run"}
|
| 394 |
+
},
|
| 395 |
+
"required": ["command"],
|
| 396 |
+
},
|
| 397 |
+
},
|
| 398 |
+
}
|
| 399 |
+
]
|
| 400 |
+
|
| 401 |
+
messages = [{"role": "user", "content": "List the Python files in the current directory."}]
|
| 402 |
+
|
| 403 |
+
response = client.chat.completions.create(
|
| 404 |
+
model="deepreinforce-ai/Ornith-1.5-9B",
|
| 405 |
+
messages=messages,
|
| 406 |
+
tools=tools,
|
| 407 |
+
temperature=0.6,
|
| 408 |
+
top_p=0.95,
|
| 409 |
+
)
|
| 410 |
+
print(response.choices[0].message)
|
| 411 |
+
```
|
| 412 |
+
|
| 413 |
+
**Examples of using Ornith with agent harness:**
|
| 414 |
+
|
| 415 |
+
#### Hermes Agent
|
| 416 |
+
```bash
|
| 417 |
+
# Hermes talks to any OpenAI-compatible endpoint — point it at your Ornith server.
|
| 418 |
+
export OPENAI_BASE_URL="http://localhost:8000/v1"
|
| 419 |
+
export OPENAI_API_KEY="EMPTY"
|
| 420 |
+
export MODEL="deepreinforce-ai/Ornith-1.5-9B"
|
| 421 |
+
```
|
| 422 |
+
|
| 423 |
+
|
| 424 |
+
#### Atomic.chat / Ollama / llama.cpp
|
| 425 |
+
```bash
|
| 426 |
+
# Both runtimes load a GGUF build of Ornith (publish one at deepreinforce-ai/Ornith-1.5-9B-GGUF).
|
| 427 |
+
|
| 428 |
+
# llama.cpp — serve an OpenAI-compatible API on port 8000.
|
| 429 |
+
llama-server -hf deepreinforce-ai/Ornith-1.5-9B-GGUF --port 8000 -c 262144
|
| 430 |
+
|
| 431 |
+
# Ollama — pull and chat with the same GGUF straight from Hugging Face.
|
| 432 |
+
ollama run hf.co/deepreinforce-ai/Ornith-1.5-9B-GGUF
|
| 433 |
+
```
|
| 434 |
+
|
| 435 |
+
#### OpenClaw
|
| 436 |
+
|
| 437 |
+
```bash
|
| 438 |
+
# OpenClaw talks to any OpenAI-compatible endpoint — point it at your Ornith server.
|
| 439 |
+
export OPENAI_BASE_URL="http://localhost:8000/v1"
|
| 440 |
+
export OPENAI_API_KEY="EMPTY"
|
| 441 |
+
export OPENAI_MODEL="deepreinforce-ai/Ornith-1.5-9B"
|
| 442 |
+
```
|
| 443 |
+
|
| 444 |
+
#### Unsloth Studio
|
| 445 |
+
|
| 446 |
+
```bash
|
| 447 |
+
pip install unsloth
|
| 448 |
+
|
| 449 |
+
# Load Ornith for fast local inference or fine-tuning (Python):
|
| 450 |
+
# from unsloth import FastLanguageModel
|
| 451 |
+
# model, tokenizer = FastLanguageModel.from_pretrained(
|
| 452 |
+
# "deepreinforce-ai/Ornith-1.5-9B",
|
| 453 |
+
# max_seq_length=262144,
|
| 454 |
+
# load_in_4bit=True,
|
| 455 |
+
# )
|
| 456 |
+
```
|
| 457 |
+
|
| 458 |
+
#### OpenHands
|
| 459 |
+
```bash
|
| 460 |
+
pip install openhands-ai
|
| 461 |
+
|
| 462 |
+
# OpenHands routes through LiteLLM; the "openai/" prefix selects the OpenAI-compatible path.
|
| 463 |
+
export LLM_MODEL="openai/deepreinforce-ai/Ornith-1.5-9B"
|
| 464 |
+
export LLM_BASE_URL="http://localhost:8000/v1"
|
| 465 |
+
export LLM_API_KEY="EMPTY"
|
| 466 |
+
|
| 467 |
+
# Launch the CLI (or run the official OpenHands Docker image with the same env vars).
|
| 468 |
+
openhands
|
| 469 |
+
```
|
| 470 |
+
|
| 471 |
+
### Coding CLIs
|
| 472 |
+
|
| 473 |
+
Ornith-1.5-9B is optimized for terminal-based coding agents. Point any OpenAI-compatible coding CLI at your Ornith-1.5-9B endpoint (set `OPENAI_BASE_URL` and `OPENAI_API_KEY`) to understand large codebases, automate tedious work, and ship faster.
|
| 474 |
+
|
| 475 |
+
#### OpenCode
|
| 476 |
+
```bash
|
| 477 |
+
# Register your local Ornith endpoint as a provider in ~/.config/opencode/opencode.json:
|
| 478 |
+
#
|
| 479 |
+
# {
|
| 480 |
+
# "$schema": "https://opencode.ai/config.json",
|
| 481 |
+
# "provider": {
|
| 482 |
+
# "ornith": {
|
| 483 |
+
# "npm": "@ai-sdk/openai-compatible",
|
| 484 |
+
# "name": "Ornith (local)",
|
| 485 |
+
# "options": { "baseURL": "http://localhost:8000/v1", "apiKey": "EMPTY" },
|
| 486 |
+
# "models": { "deepreinforce-ai/Ornith-1.5-9B": { "name": "Ornith-1.5-9B" } }
|
| 487 |
+
# }
|
| 488 |
+
# }
|
| 489 |
+
# }
|
| 490 |
+
|
| 491 |
+
opencode
|
| 492 |
+
```
|
| 493 |
+
|
| 494 |
+
### Citation
|
| 495 |
+
|
| 496 |
+
If you find our work helpful, feel free to give us a cite.
|
| 497 |
+
|
| 498 |
+
```bibtex
|
| 499 |
+
@misc{ornith_1_5,
|
| 500 |
+
title = {{Ornith-1.5}: From Self-Scaffolding to Self-Improvement},
|
| 501 |
+
url = {https://deep-reinforce.com/ornith.html},
|
| 502 |
+
author = {{Ornith Team}},
|
| 503 |
+
year = {2026}
|
| 504 |
+
}
|
| 505 |
+
```
|