DeepSeek-V4.1-Flash-1017 with about a third less thinking, and the same scores.
DeepSeek-V4.1-Flash-1017 is a stronger and already more concise model than the original V4.1-Flash, but its thinking still carries a lot of hesitation: "Wait…", "Hmm, maybe…", re-deriving answers it already had. We optimised the model's weights against that habit, while leaving alone the checking it does when it works as an agent and on hard maths. AZ-Focus still reasons step by step. It just doesn't loop.
This is the 1017 version of our DeepSeek-V4.1-Flash-AZ-Focus. If you run the original V4.1-Flash, use that one; if you run 1017, use this one.
Made by AxonZeta, an independent AI lab in Manchester, UK.
| Benchmark | DeepSeek-V4.1-Flash-1017 | AZ-Focus |
|---|---|---|
| SWE-bench Verified, first 100 tasks (mini-SWE-agent, two runs) | 94 / 94 | 94 / 93 |
Thinking on, DeepSeek's recommended sampling, mini-SWE-agent with a 100-step limit, graded with the official harness. No empty patches in any run.
| Benchmark | DeepSeek-V4.1-Flash-1017 | AZ-Focus | Thinking tokens |
|---|---|---|---|
| GPQA Diamond (198, 2 runs) | 91.4% · 6,629 tokens | 90.9% · 3,777 tokens | 43% less |
| BIG-Bench Hard (540) | 82.0% · 709 | 84.1% · 359 | 49% less |
| MMLU-Pro (500, 14 subjects) | 89.2% · 1,445 | 88.8% · 953 | 34% less |
| HumanEval (164) | 98.2% · 891 | 98.8% · 552 | 38% less |
| MATH-500 + GSM8K, held-out (450, 3 runs) | 95.0% | 95.0% | 25% less (MATH), 32% less (GSM8K) |
| AIME 2025 (30, 4 runs) | 97.5% · 6,188 | 95.8% · 4,997 | 19% less |
Accuracy and mean thinking tokens per item, same questions, same serving setup. Thinking on, DeepSeek's recommended sampling (temperature 1.0, top-p 0.95), served with SGLang. GPQA and AIME were run with a 128k-token output budget; no AZ-Focus answer came close to it.
DeepSeek-V4.1-Flash-1017 has a reasoning-effort dial from 1 to 100 (default 75), so the fair comparison is at matched settings. One run each:
BIG-Bench Hard (540 tasks)
| Effort | DeepSeek-V4.1-Flash-1017 | AZ-Focus | Thinking tokens |
|---|---|---|---|
| 75 (default) | 82.0% · 709 tokens | 84.1% · 359 tokens | 49% less |
| 50 | 82.8% · 464 | 83.3% · 278 | 40% less |
| 35 | 85.6% · 447 | 81.7% · 242 | 46% less |
| 25 | 83.7% · 300 | 82.8% · 219 | 27% less |
MMLU-Pro (500 questions, 14 subjects)
| Effort | DeepSeek-V4.1-Flash-1017 | AZ-Focus | Thinking tokens |
|---|---|---|---|
| 75 (default) | 89.2% · 1,445 tokens | 88.8% · 953 tokens | 34% less |
| 50 | 88.4% · 1,245 | 87.8% · 799 | 36% less |
| 35 | 87.8% · 1,025 | 89.0% · 657 | 36% less |
| 25 | 88.2% · 704 | 88.2% · 389 | 45% less |
Single runs on a few hundred questions move by a point or two either way. Averaged across the dial, AZ-Focus is level with the original on MMLU-Pro and about half a point lower on BIG-Bench Hard. At effort 25, AZ-Focus answers MMLU-Pro with 45% less thinking than the original at the same setting, at the same accuracy.
Use it exactly like deepseek-ai/DeepSeek-V4.1-Flash-1017:
same serving engines, parallelism, chat template and recommended sampling. The difference shows in thinking mode, and it
stacks with the reasoning-effort setting (chat_template_kwargs: {"reasoning_effort": N} in SGLang). With thinking off,
output length is unchanged: the edit shortens deliberation, not answers.
MIT, as for the original model. Based on
DeepSeek-V4.1-Flash-1017 by DeepSeek-AI (see LICENSE).
Questions, or want to build on this with us? hello@axonzeta.ai
model-00048-of-00048.safetensors 102 GBmodel-00047-of-00048.safetensors 102 GBmodel-00017-of-00048.safetensors 7.4 GBmodel-00005-of-00048.safetensors 7.4 GBmodel-00011-of-00048.safetensors 7.4 GBmodel-00023-of-00048.safetensors 7.4 GBmodel-00027-of-00048.safetensors 7.4 GBmodel-00031-of-00048.safetensors 7.4 GBmodel-00035-of-00048.safetensors 7.4 GBmodel-00039-of-00048.safetensors 7.4 GBmodel-00026-of-00048.safetensors 7.4 GBmodel-00028-of-00048.safetensors 7.4 GBmodel-00029-of-00048.safetensors 7.4 GBmodel-00030-of-00048.safetensors 7.4 GBmodel-00032-of-00048.safetensors 7.4 GBmodel-00033-of-00048.safetensors 7.4 GBmodel-00034-of-00048.safetensors 7.4 GBmodel-00036-of-00048.safetensors 7.4 GBmodel-00037-of-00048.safetensors 7.4 GBmodel-00038-of-00048.safetensors 7.4 GBmodel-00013-of-00048.safetensors 7.4 GBmodel-00014-of-00048.safetensors 7.4 GBmodel-00015-of-00048.safetensors 7.4 GBmodel-00016-of-00048.safetensors 7.4 GBmodel-00018-of-00048.safetensors 7.4 GBmodel-00019-of-00048.safetensors 7.4 GBmodel-00020-of-00048.safetensors 7.4 GBmodel-00021-of-00048.safetensors 7.4 GBmodel-00022-of-00048.safetensors 7.4 GBmodel-00024-of-00048.safetensors 7.4 GBmodel-00025-of-00048.safetensors 7.4 GBmodel-00040-of-00048.safetensors 7.4 GBmodel-00041-of-00048.safetensors 7.4 GBmodel-00042-of-00048.safetensors 7.4 GBmodel-00003-of-00048.safetensors 7.4 GBmodel-00004-of-00048.safetensors 7.4 GBmodel-00006-of-00048.safetensors 7.4 GBmodel-00007-of-00048.safetensors 7.4 GBmodel-00008-of-00048.safetensors 7.4 GBmodel-00009-of-00048.safetensors 7.4 GBmodel-00010-of-00048.safetensors 7.4 GBmodel-00012-of-00048.safetensors 7.4 GBmodel-00046-of-00048.safetensors 2.7 GBmodel-00044-of-00048.safetensors 2.7 GBmodel-00045-of-00048.safetensors 2.6 GBmodel-00002-of-00048.safetensors 1.3 GBmodel-00043-of-00048.safetensors 1.3 GBmodel-00001-of-00048.safetensors 971 MB