DeepSeek V4.1 Flash
deepseek-flash
DeepSeek V4.1 Flash model for reasoning and agentic coding
ReasoningTool CallingAttachmentsStructured OutputOpen WeightsMIT
Context Window
1M
Max Output
384K
Temperature
Yes
Open Weights
Yes
Knowledge Cutoff
2025-05
Released
2026-09-10
Last Updated
2026-09-10
License
MIT
Modalities
Input:TextImage
→Output:Text
Benchmarks
| Benchmark | Score | Metric | Harness |
|---|---|---|---|
| GPQA Diamond | 90.9 | pass@1 | - |
| MathArena Apex | 65.6 | pass@1 | - |
| Humanity's Last Exam | 36.8 | pass@1 | - |
| Humanity's Last Exam | 39.1 | pass@1 | - |
| Codeforces | 3471 | rating | - |
| Terminal-Bench | 90.6 | pass@1 | DeepSeek Harness minimal mode |
| Terminal-Bench | 30 | pass@1 | DeepSeek Harness minimal mode |
| Terminal-Bench | 31.2 | pass@1 | DeepSeek Harness minimal mode |
| DeepSWE | 74.2 | resolved | mini-swe-agent |
| ProgramBench | 20.3 | almost@1 | DeepSeek Harness minimal mode |
| NL2Repo | 64 | score | DeepSeek Harness minimal mode |
| CyberGym | 88.1 | pass@1 | - |
| ExploitGym | 15.3 | pass@1 | - |
| AutomationBench | 54.8 | pass@1 | - |
| Agents' Last Exam | 31.8 | pass@1 | - |
| SEC-Bench Pro | 62.8 | pass@1 | Claude Code |
| Humanity's Last Exam | 63.9 | pass@1 | - |
| Chartography | 78.9 | pass@1 | Claude Code |
| BabyVision | 89.6 | pass@1 | Claude Code |
| ZeroBench | 49 | pass@5 | Claude Code |
| DeepSWE | 69.8 | resolved | Claude Code |
| Terminal-Bench | 88 | pass@1 | Claude Code |
| DeepSWE | 65.6 | resolved | Codex |
| Terminal-Bench | 84.1 | pass@1 | Codex |
| DeepSWE | 65.5 | resolved | OpenCode |
| Terminal-Bench | 85 | pass@1 | OpenCode |
| DeepSWE | 66.2 | resolved | Pi |
| Terminal-Bench | 86.1 | pass@1 | Pi |
| Terminal-Bench | 90.3 | pass@1 | mini-swe-agent |
| DeepSWE | 72.6 | resolved | DeepSeek Harness minimal mode |
| DeepSWE | 70.5 | resolved | DeepSeek Harness standard mode |
| Terminal-Bench | 85.8 | pass@1 | DeepSeek Harness standard mode |
| DeepSWE | 67.6 | resolved | DeepSeek Harness PTC mode |
| Terminal-Bench | 85.8 | pass@1 | DeepSeek Harness PTC mode |
Available Providers (11)
| Provider | Input /1M | Output /1M | Cache Read /1M | Cache Write /1M | Reasoning | Status |
|---|---|---|---|---|---|---|
 NanoGPT | Rp 2.325 | Rp 9.298 | Rp 108 | — | effort | — |
 LLM Gateway | Rp 2.683 | Rp 10.729 | Rp 54 | — | effort | — |
 Kilo Gateway | Rp 5.365 | Rp 21.457 | Rp 108 | — | effort | — |
 Novita AI | Rp 5.365 | Rp 21.457 | Rp 108 | — | toggle, effort | — |
 OpenRouter | Rp 5.365 | Rp 21.457 | Rp 108 | — | toggle, effort | — |
 ZenMux | Rp 2.683 | Rp 10.729 | Rp 54 | — | toggle, effort | — |
 Vercel AI Gateway | Rp 5.365 | Rp 21.457 | Rp 126 | — | effort | — |
 Ofox | Rp 3.755 | Rp 15.020 | Rp 76 | — | toggle, effort | — |
 Merge Gateway | Rp 2.683 | Rp 10.729 | Rp 54 | — | toggle, effort | — |
 TensorX | Rp 8.941 | Rp 26.821 | Rp 2.325 | — | toggle | — |
 CrossModel | Rp 5.365 | Rp 21.457 | Rp 108 | Rp 5.365 | toggle, effort | — |