volkermauel/nebius-swe-rebench-openhands-qwen3-30b-a3b-gguf overview
SWE-rebench-openhands-Qwen3-30B-A3B is a 30B Rejection Sampling Fine-Tuning (RFT) checkpoint derived from Qwen/Qwen3-30B-A3B-Instruct-2507, trained on the newly released nebius/SWE-rebench-openhands-trajectories dataset. Training used a maximum sequence length of 131k tokens. Model Size Maximum Number of Turns = 100 Maximum Number of Turns = 500 Pass@1 Pass@5 Pass@1 Pass@5 Pass@1 Pass@5 Pass@1 Pass@5 30B scale Qwen3-30B-A3B-Instruct-2507 30B 25.2 44.8 11.8 24.4 25.7 44.2 14.2 26.5 Qwen3-Coder-30B-A3B-Instruct 30B 51.9 67.3 28.7 42.8 50.0 63.0 28.1 38.7 nebius/SWE-rebench-openhands-Qwen3-30B-A3B (Ours) 30B 49.7(+24.5) 65.4(+20.6) 28.1(+16.3) 38.7(+14.3) 50.3(+24.6) 68.3(+24.1) 28.1(+13.9) 38.7(+12.2) 100B+ scale GLM-4.5-Air 106B 58.2 73.5 33.8 42.8 - - - - 200B+ scale Qwen3-235B-A22B-Instruct-2507 235B 45.2 65.9 29.3 44.8 46.2 67.5 25.3 40.8 nebius/SWE-rebench-openhands-Qwen3-235B-A22B (Ours) 235B 59.9(+14.7) 73.9(+8.0) 35.1(+5.8) 46.9(+2.1) 61.7(+15.5) 74.3(+6.8) 34.2(+8.9) 44.8(+4.0) 300B+ scale GLM-4.5 355B 64.4 76.2 33.8 44.8 - - - - Qwen3-Coder-480B-A35B-Instruct 480B 64.7 75.8 36.3 44.8 66.5 77.8 35.5 42.8 Table 1. Pass@1 (averaged over 5 runs) and Pass@5 for OpenHands agent with the maximum number of turns set to 100 (highlighted in yellow) and 500 (highlighted in green). Metrics are reported in percentages. Deltas vs base models are shown in parentheses for fine-tuned models. We explicitly excluded all SWE-bench Verified and SWE-rebench September issues from training to avoid contamination. SWE-rebench Verified was additionally decontaminated on repository level. When evaluated with the OpenHands (v0.54.0) agent, our 30B model: Substantially improves over the base Qwen3-30B-A3B-Instruct-2507 model, with +24.6 Pass@1 and +24.1 Pass@5 gains at 500-turn settings. Matches or surpasses the specialized Qwen3-Coder-30B-A3B-Instruct baseline at 500 turns on SWE-bench Verified (50.3% vs 50.0% Pass@1; 68.3% vs 63.0% Pass@5). Generalizes to longer interaction horizons, despite training on trajectories capped at 100 turns. For more details see our report in Nebius blog. --- # Best Practices 1. Deployment: Use the following configuration to serve the model with vLLM: Tested using vllm/vllm-openai:v0.9.0 Docker image. 2. Sampling Parameters: * For optimal performance, we recommend Temperature=0.7, TopP=0.8, TopK=20, and MinP=0 that are consistent with the base model. --- # Citation
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| imatrix.gguf | GGUF | — | 116.38 MB | Download |
| model-f16.gguf | GGUF | F16 | 56.90 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ1_M.gguf | GGUF | IQ1_M | 6.97 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ1_S.gguf | GGUF | IQ1_S | 6.36 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ2_M.gguf | GGUF | IQ2_M | 9.85 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ2_S.gguf | GGUF | IQ2_S | 9.03 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ2_XS.gguf | GGUF | IQ2_XS | 8.83 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ2_XXS.gguf | GGUF | IQ2_XXS | 8.00 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ3_M.gguf | GGUF | IQ3_M | 12.93 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ3_S.gguf | GGUF | IQ3_S | 12.73 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ3_XS.gguf | GGUF | IQ3_XS | 12.08 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ3_XXS.gguf | GGUF | IQ3_XXS | 11.42 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ4_NL.gguf | GGUF | IQ4_NL | 16.46 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-IQ4_XS.gguf | GGUF | IQ4_XS | 15.59 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-MXFP4_MOE.gguf | GGUF | — | 16.18 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q2_K.gguf | GGUF | Q2_K | 10.83 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q2_K_S.gguf | GGUF | Q2_K_S | 10.14 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q3_K.gguf | GGUF | Q3_K | 14.04 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q3_K_L.gguf | GGUF | Q3_K_L | 15.15 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q3_K_M.gguf | GGUF | Q3_K_M | 14.04 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q3_K_S.gguf | GGUF | Q3_K_S | 12.72 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q4_K.gguf | GGUF | Q4_K | 17.62 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q4_K_M.gguf | GGUF | Q4_K_M | 17.62 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q4_K_S.gguf | GGUF | Q4_K_S | 16.60 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q5_K.gguf | GGUF | Q5_K | 20.58 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q5_K_M.gguf | GGUF | Q5_K_M | 20.58 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q5_K_S.gguf | GGUF | Q5_K_S | 19.97 GB | Download |
| nebius-SWE-rebench-openhands-Qwen3-30B-A3B-Q6_K.gguf | GGUF | Q6_K | 23.71 GB | Download |
Benchmarks
| Model | Size | Maximum Number of Turns = 100 | Maximum Number of Turns = 500 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Pass@1 | Pass@5 | Pass@1 | Pass@5 | Pass@1 | Pass@5 | Pass@1 | Pass@5 | ||
| 30B scale | |||||||||
| Qwen3-30B-A3B-Instruct-2507 | 30B | 25.2 | 44.8 | 11.8 | 24.4 | 25.7 | 44.2 | 14.2 | 26.5 |
| Qwen3-Coder-30B-A3B-Instruct | 30B | 51.9 | 67.3 | 28.7 | 42.8 | 50.0 | 63.0 | 28.1 | 38.7 |
| nebius/SWE-rebench-openhands-Qwen3-30B-A3B (Ours) | 30B | 49.7 (+24.5) |
65.4 (+20.6) |
28.1 (+16.3) |
38.7 (+14.3) |
50.3 (+24.6) |
68.3 (+24.1) |
28.1 (+13.9) |
38.7 (+12.2) |
| 100B+ scale | |||||||||
| GLM-4.5-Air | 106B | 58.2 | 73.5 | 33.8 | 42.8 | - | - | - | - |
| 200B+ scale | |||||||||
| Qwen3-235B-A22B-Instruct-2507 | 235B | 45.2 | 65.9 | 29.3 | 44.8 | 46.2 | 67.5 | 25.3 | 40.8 |
| nebius/SWE-rebench-openhands-Qwen3-235B-A22B (Ours) | 235B | 59.9 (+14.7) |
73.9 (+8.0) |
35.1 (+5.8) |
46.9 (+2.1) |
61.7 (+15.5) |
74.3 (+6.8) |
34.2 (+8.9) |
44.8 (+4.0) |
| 300B+ scale | |||||||||
| GLM-4.5 | 355B | 64.4 | 76.2 | 33.8 | 44.8 | - | - | - | - |
| Qwen3-Coder-480B-A35B-Instruct | 480B | 64.7 | 75.8 | 36.3 | 44.8 | 66.5 | 77.8 | 35.5 | 42.8 |
Model Details Live
Metadata Inspector
Normalized metadata (stored in metadata_json)
{
"metadata": {},
"card_data": {
"license": "apache-2.0",
"datasets": [
"nebius/SWE-rebench-openhands-trajectories"
],
"base_model": [
"nebius/SWE-rebench-openhands-Qwen3-30B-A3B"
],
"pipeline_tag": "text-generation",
"library_name": "transformers",
"tags": [
"code",
"agent"
],
"frontmatter": {
"license": "apache-2.0",
"datasets": [
"nebius/SWE-rebench-openhands-trajectories"
],
"base_model": [
"nebius/SWE-rebench-openhands-Qwen3-30B-A3B"
],
"pipeline_tag": "text-generation",
"library_name": "transformers",
"tags": [
"code",
"agent"
]
},
"hero_image_url": "",
"summary": "**SWE-rebench-openhands-Qwen3-30B-A3B** is a 30B Rejection Sampling Fine-Tuning (RFT) checkpoint derived from Qwen/Qwen3-30B-A3B-Instruct-2507, trained on the newly released nebius/SWE-rebench-openhands-trajectories dataset. Training used a maximum sequence length of 131k tokens. Model Size Maximum Number of Turns = 100 Maximum Number of Turns = 500 Pass@1 Pass@5 Pass@1 Pass@5 Pass@1 Pass@5 Pass@1 Pass@5 30B scale Qwen3-30B-A3B-Instruct-2507 30B 25.2 44.8 11.8 24.4 25.7 44.2 14.2 26.5 Qwen3-Coder-30B-A3B-Instruct 30B 51.9 67.3 28.7 42.8 50.0 63.0 28.1 38.7 nebius/SWE-rebench-openhands-Qwen3-30B-A3B (Ours) 30B 49.7(+24.5) 65.4(+20.6) 28.1(+16.3) 38.7(+14.3) 50.3(+24.6) 68.3(+24.1) 28.1(+13.9) 38.7(+12.2) 100B+ scale GLM-4.5-Air 106B 58.2 73.5 33.8 42.8 - - - - 200B+ scale Qwen3-235B-A22B-Instruct-2507 235B 45.2 65.9 29.3 44.8 46.2 67.5 25.3 40.8 nebius/SWE-rebench-openhands-Qwen3-235B-A22B (Ours) 235B 59.9(+14.7) 73.9(+8.0) 35.1(+5.8) 46.9(+2.1) 61.7(+15.5) 74.3(+6.8) 34.2(+8.9) 44.8(+4.0) 300B+ scale GLM-4.5 355B 64.4 76.2 33.8 44.8 - - - - Qwen3-Coder-480B-A35B-Instruct 480B 64.7 75.8 36.3 44.8 66.5 77.8 35.5 42.8 **Table 1.** Pass@1 (averaged over 5 runs) and Pass@5 for OpenHands agent with the maximum number of turns set to 100 (highlighted in yellow) and 500 (highlighted in green). Metrics are reported in percentages. Deltas vs base models are shown in parentheses for fine-tuned models. We explicitly excluded all SWE-bench Verified and SWE-rebench September issues from training to avoid contamination. SWE-rebench Verified was additionally decontaminated on repository level. When evaluated with the OpenHands (v0.54.0) agent, our 30B model: * Substantially improves over the base Qwen3-30B-A3B-Instruct-2507 model, with **+24.6 Pass@1** and **+24.1 Pass@5** gains at 500-turn settings. * Matches or surpasses the specialized Qwen3-Coder-30B-A3B-Instruct baseline at 500 turns on SWE-bench Verified (**50.3%** vs **50.0% Pass@1**; **68.3%** vs **63.0% Pass@5**). * Generalizes to longer interaction horizons, despite training on trajectories capped at 100 turns. For more details see our report in Nebius blog. --- # Best Practices 1. **Deployment:** * Use the following configuration to serve the model with vLLM: ``bash VLLM_USE_V1=1 vllm serve nebius/SWE-rebench-openhands-Qwen3-30B-A3B --tensor-parallel-size 8 --served-model-name qwen_3_instruct_2507 --disable-log-requests --enable-prefix-caching --max-model-len 131072 --enable-auto-tool-choice --tool-call-parser hermes ` Tested using vllm/vllm-openai:v0.9.0 Docker image. 2. **Sampling Parameters:** * For optimal performance, we recommend Temperature=0.7, TopP=0.8, TopK=20, and MinP=0 that are consistent with the base model. --- # Citation ` @article{trofimova2025openhandstrajs, title={OpenHands Trajectories with Qwen3-Coder-480B-A35B-Instruct}, author={Trofimova, Maria and Shevtsov, Anton and Ibragim, Badertdinov and Pyaev, Konstantin and Karasik, Simon and Golubev, Alexander}, year={2025}, journal={Nebius blog}, note={} } ``",
"quick_links": [],
"benchmark_table_html": "<table>\n <thead>\n <tr>\n <th rowspan=\"2\">Model</th>\n <th rowspan=\"2\">Size</th>\n <th colspan=\"4\" style=\"background-color: #fff3cd;\">Maximum Number of Turns = 100</th>\n <th colspan=\"4\" style=\"background-color: #d4edda;\">Maximum Number of Turns = 500</th>\n </tr>\n <tr>\n <th style=\"background-color: #fff3cd;\">Pass@1</th>\n <th style=\"background-color: #fff3cd;\">Pass@5</th>\n <th style=\"background-color: #fff3cd;\">Pass@1</th>\n <th style=\"background-color: #fff3cd;\">Pass@5</th>\n <th style=\"background-color: #d4edda;\">Pass@1</th>\n <th style=\"background-color: #d4edda;\">Pass@5</th>\n <th style=\"background-color: #d4edda;\">Pass@1</th>\n <th style=\"background-color: #d4edda;\">Pass@5</th>\n </tr>\n </thead>\n <tbody>\n <tr>\n <td colspan=\"10\"><strong>30B scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507\">Qwen3-30B-A3B-Instruct-2507</a></td>\n <td>30B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">25.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">11.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">24.4</td>\n <td style=\"background-color: #d4edda;text-align: center;\">25.7</td>\n <td style=\"background-color: #d4edda;text-align: center;\">44.2</td>\n <td style=\"background-color: #d4edda;text-align: center;\">14.2</td>\n <td style=\"background-color: #d4edda;text-align: center;\">26.5</td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct\">Qwen3-Coder-30B-A3B-Instruct</a></td>\n <td>30B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>51.9</strong></td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>67.3</strong></td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>28.7</strong></td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>42.8</strong></td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>50.0</strong></td>\n <td style=\"background-color: #d4edda;text-align: center;\">63.0</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>28.1</strong></td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>38.7</strong></td>\n </tr>\n <tr style=\"background-color: #ebeced\">\n <td style=\"color: black;\">nebius/SWE-rebench-openhands-Qwen3-30B-A3B (Ours)</td>\n <td>30B</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">49.7<br/>(+24.5)</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">65.4<br/>(+20.6)</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">28.1<br/>(+16.3)</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">38.7<br/>(+14.3)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>50.3</strong><br/>(+24.6)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>68.3</strong><br/>(+24.1)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>28.1</strong><br/>(+13.9)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>38.7</strong><br/>(+12.2)</td>\n </tr>\n <tr>\n <td colspan=\"10\"><strong>100B+ scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/zai-org/GLM-4.5-Air\">GLM-4.5-Air</a></td>\n <td>106B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">58.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">73.5</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">33.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">42.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n </tr>\n <tr>\n <td colspan=\"10\"><strong>200B+ scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507\">Qwen3-235B-A22B-Instruct-2507</a></td>\n <td>235B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">45.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">65.9</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">29.3</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">46.2</td>\n <td style=\"background-color: #d4edda;text-align: center;\">67.5</td>\n <td style=\"background-color: #d4edda;text-align: center;\">25.3</td>\n <td style=\"background-color: #d4edda;text-align: center;\">40.8</td>\n </tr>\n <tr>\n <td style=\"color: black;\"><a href=\"https://huggingface.co/nebius/SWE-rebench-openhands-Qwen3-235B-A22B\">nebius/SWE-rebench-openhands-Qwen3-235B-A22B</a> (Ours)</td>\n <td>235B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>59.9</strong><br/>(+14.7)</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>73.9</strong><br/>(+8.0)</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>35.1</strong><br/>(+5.8)</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>46.9</strong><br/>(+2.1)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>61.7</strong><br/>(+15.5)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>74.3</strong><br/>(+6.8)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>34.2</strong><br/>(+8.9)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>44.8</strong><br/>(+4.0)</td>\n </tr>\n <tr>\n <td colspan=\"10\"><strong>300B+ scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/zai-org/GLM-4.5\">GLM-4.5</a></td>\n <td>355B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">64.4</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">76.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">33.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-Coder-480B-A35B-Instruct\">Qwen3-Coder-480B-A35B-Instruct</a></td>\n <td>480B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">64.7</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">75.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">36.3</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">66.5</td>\n <td style=\"background-color: #d4edda;text-align: center;\">77.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">35.5</td>\n <td style=\"background-color: #d4edda;text-align: center;\">42.8</td>\n </tr>\n </tbody>\n</table>",
"readme_markdown": "---\nlicense: apache-2.0\ndatasets:\n- nebius/SWE-rebench-openhands-trajectories\nbase_model:\n- nebius/SWE-rebench-openhands-Qwen3-30B-A3B\npipeline_tag: text-generation\nlibrary_name: transformers\ntags:\n- code\n- agent\n---\n\nConverted with imatrix from calib created via codesearchnet.\n\n# Model Summary\n\n**SWE-rebench-openhands-Qwen3-30B-A3B** is a 30B Rejection Sampling Fine-Tuning (RFT) checkpoint derived from\n[Qwen/Qwen3-30B-A3B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507), trained on the newly released\n[nebius/SWE-rebench-openhands-trajectories](https://huggingface.co/datasets/nebius/SWE-rebench-openhands-trajectories) dataset.\nTraining used a maximum sequence length of 131k tokens.\n\n<table>\n <thead>\n <tr>\n <th rowspan=\"2\">Model</th>\n <th rowspan=\"2\">Size</th>\n <th colspan=\"4\" style=\"background-color: #fff3cd;\">Maximum Number of Turns = 100</th>\n <th colspan=\"4\" style=\"background-color: #d4edda;\">Maximum Number of Turns = 500</th>\n </tr>\n <tr>\n <th style=\"background-color: #fff3cd;\">Pass@1</th>\n <th style=\"background-color: #fff3cd;\">Pass@5</th>\n <th style=\"background-color: #fff3cd;\">Pass@1</th>\n <th style=\"background-color: #fff3cd;\">Pass@5</th>\n <th style=\"background-color: #d4edda;\">Pass@1</th>\n <th style=\"background-color: #d4edda;\">Pass@5</th>\n <th style=\"background-color: #d4edda;\">Pass@1</th>\n <th style=\"background-color: #d4edda;\">Pass@5</th>\n </tr>\n </thead>\n <tbody>\n <tr>\n <td colspan=\"10\"><strong>30B scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507\">Qwen3-30B-A3B-Instruct-2507</a></td>\n <td>30B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">25.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">11.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">24.4</td>\n <td style=\"background-color: #d4edda;text-align: center;\">25.7</td>\n <td style=\"background-color: #d4edda;text-align: center;\">44.2</td>\n <td style=\"background-color: #d4edda;text-align: center;\">14.2</td>\n <td style=\"background-color: #d4edda;text-align: center;\">26.5</td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct\">Qwen3-Coder-30B-A3B-Instruct</a></td>\n <td>30B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>51.9</strong></td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>67.3</strong></td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>28.7</strong></td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>42.8</strong></td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>50.0</strong></td>\n <td style=\"background-color: #d4edda;text-align: center;\">63.0</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>28.1</strong></td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>38.7</strong></td>\n </tr>\n <tr style=\"background-color: #ebeced\">\n <td style=\"color: black;\">nebius/SWE-rebench-openhands-Qwen3-30B-A3B (Ours)</td>\n <td>30B</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">49.7<br/>(+24.5)</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">65.4<br/>(+20.6)</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">28.1<br/>(+16.3)</td>\n <td style=\"background-color: #ffdf80;text-align: center;\">38.7<br/>(+14.3)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>50.3</strong><br/>(+24.6)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>68.3</strong><br/>(+24.1)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>28.1</strong><br/>(+13.9)</td>\n <td style=\"background-color: #9df2b3;text-align: center;\"><strong>38.7</strong><br/>(+12.2)</td>\n </tr>\n <tr>\n <td colspan=\"10\"><strong>100B+ scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/zai-org/GLM-4.5-Air\">GLM-4.5-Air</a></td>\n <td>106B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">58.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">73.5</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">33.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">42.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n </tr>\n <tr>\n <td colspan=\"10\"><strong>200B+ scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507\">Qwen3-235B-A22B-Instruct-2507</a></td>\n <td>235B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">45.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">65.9</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">29.3</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">46.2</td>\n <td style=\"background-color: #d4edda;text-align: center;\">67.5</td>\n <td style=\"background-color: #d4edda;text-align: center;\">25.3</td>\n <td style=\"background-color: #d4edda;text-align: center;\">40.8</td>\n </tr>\n <tr>\n <td style=\"color: black;\"><a href=\"https://huggingface.co/nebius/SWE-rebench-openhands-Qwen3-235B-A22B\">nebius/SWE-rebench-openhands-Qwen3-235B-A22B</a> (Ours)</td>\n <td>235B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>59.9</strong><br/>(+14.7)</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>73.9</strong><br/>(+8.0)</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>35.1</strong><br/>(+5.8)</td>\n <td style=\"background-color: #fff3cd;text-align: center;\"><strong>46.9</strong><br/>(+2.1)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>61.7</strong><br/>(+15.5)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>74.3</strong><br/>(+6.8)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>34.2</strong><br/>(+8.9)</td>\n <td style=\"background-color: #d4edda;text-align: center;\"><strong>44.8</strong><br/>(+4.0)</td>\n </tr>\n <tr>\n <td colspan=\"10\"><strong>300B+ scale</strong></td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/zai-org/GLM-4.5\">GLM-4.5</a></td>\n <td>355B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">64.4</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">76.2</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">33.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n <td style=\"background-color: #d4edda;text-align: center;\">-</td>\n </tr>\n <tr>\n <td><a href=\"https://huggingface.co/Qwen/Qwen3-Coder-480B-A35B-Instruct\">Qwen3-Coder-480B-A35B-Instruct</a></td>\n <td>480B</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">64.7</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">75.8</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">36.3</td>\n <td style=\"background-color: #fff3cd;text-align: center;\">44.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">66.5</td>\n <td style=\"background-color: #d4edda;text-align: center;\">77.8</td>\n <td style=\"background-color: #d4edda;text-align: center;\">35.5</td>\n <td style=\"background-color: #d4edda;text-align: center;\">42.8</td>\n </tr>\n </tbody>\n</table>\n\n**Table 1.** Pass@1 (averaged over 5 runs) and Pass@5 for OpenHands agent with the maximum number of turns set to 100\n(highlighted in <span style=\"background-color: #fff3cd; padding: 4px;\">yellow</span>) and 500\n(highlighted in <span style=\"background-color: #d4edda; padding: 4px;\">green</span>). Metrics are reported in percentages.\nDeltas vs base models are shown in parentheses for fine-tuned models.\n\nWe explicitly excluded all [SWE-bench Verified](https://huggingface.co/datasets/princeton-nlp/SWE-bench_Verified) and\n[SWE-rebench September](https://huggingface.co/datasets/nebius/SWE-rebench-leaderboard) issues from training to avoid contamination.\nSWE-rebench Verified was additionally decontaminated on repository level.\n\nWhen evaluated with the OpenHands (v0.54.0) agent, our 30B model:\n\n* Substantially improves over the base Qwen3-30B-A3B-Instruct-2507 model, with **+24.6 Pass@1** and **+24.1 Pass@5** gains at 500-turn settings.\n* Matches or surpasses the specialized Qwen3-Coder-30B-A3B-Instruct baseline at 500 turns on SWE-bench Verified (**50.3%** vs **50.0% Pass@1**; **68.3%** vs **63.0% Pass@5**).\n* Generalizes to longer interaction horizons, despite training on trajectories capped at 100 turns.\n\nFor more details see our report in [Nebius blog](https://nebius.com/blog/posts/openhands-trajectories-with-qwen3-coder-480b).\n\n---\n\n# Best Practices\n\n1. **Deployment:**\n * Use the following configuration to serve the model with vLLM:\n ```bash\n VLLM_USE_V1=1 vllm serve nebius/SWE-rebench-openhands-Qwen3-30B-A3B\n --tensor-parallel-size 8\n --served-model-name qwen_3_instruct_2507\n --disable-log-requests\n --enable-prefix-caching\n --max-model-len 131072\n --enable-auto-tool-choice\n --tool-call-parser hermes\n ```\n Tested using `vllm/vllm-openai:v0.9.0` Docker image.\n\n2. **Sampling Parameters:**\n * For optimal performance, we recommend `Temperature=0.7`, `TopP=0.8`, `TopK=20`, and `MinP=0`\n that are consistent with the base model.\n \n---\n\n# Citation\n\n```\n@article{trofimova2025openhandstrajs,\n title={OpenHands Trajectories with Qwen3-Coder-480B-A35B-Instruct},\n author={Trofimova, Maria and Shevtsov, Anton and Ibragim, Badertdinov and Pyaev, Konstantin and Karasik, Simon and Golubev, Alexander},\n year={2025},\n journal={Nebius blog},\n note={}\n}\n```\n",
"related_quantizations": []
},
"tags": [
"transformers",
"gguf",
"code",
"agent",
"text-generation",
"dataset:nebius/SWE-rebench-openhands-trajectories",
"base_model:nebius/SWE-rebench-openhands-Qwen3-30B-A3B",
"base_model:quantized:nebius/SWE-rebench-openhands-Qwen3-30B-A3B",
"license:apache-2.0",
"endpoints_compatible",
"region:us",
"conversational"
],
"likes": 2,
"downloads": 394,
"gated": false,
"private": false,
"last_modified": "2025-12-28T00:47:52.000Z",
"created_at": "2025-12-27T12:30:14.000Z",
"pipeline_tag": "text-generation",
"library_name": "transformers"
}
Source payload excerpt (from Hugging Face API)
{
"_id": "694fd156785ac06cb1e114cd",
"id": "VolkerMauel/nebius-SWE-rebench-openhands-Qwen3-30B-A3B-GGUF",
"modelId": "VolkerMauel/nebius-SWE-rebench-openhands-Qwen3-30B-A3B-GGUF",
"sha": "b42980a91b94e6b7e8a02f6e810098d64df0f223",
"createdAt": "2025-12-27T12:30:14.000Z",
"lastModified": "2025-12-28T00:47:52.000Z",
"author": "VolkerMauel",
"downloads": 394,
"likes": 2,
"gated": false,
"private": false,
"pipeline_tag": "text-generation",
"library_name": "transformers",
"siblings_count": 30
}