DeepSeek-V4.1-Flash: benchmarks, price and rank
DeepSeek-V4.1-Flash from DeepSeek is 27th of 123 rated AI models on Spezia's leaderboard, strongest on APEX-Agents and CyberGym. It costs $0.22 in, $0.66 out per million tokens.
- 27th of 123 rated models
- Rating 1,530
- Rated on 13 benchmarks
- Open weights
- Released Sep 9, 2026
- 1 million tokens of context
Where it is strongest
- 6th of 56 on APEX-Agents
- 1st of 15 on CyberGym
Where it falls behind
- 16th of 17 on Terminal-Bench 4.0
- 6th of 8 on ExploitGym
- 45th of 78 on SimpleQA Verified
Price: $0.22 in, $0.66 out per million tokens.
Its headline results
The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.
- APEX-Agents69.8%
Long tasks written by investment banking analysts, management consultants and corporate lawyers from their own work, done across files and apps in a realistic work setup.
6th of 56. Independent result from Epoch AI.
- Terminal-Bench 4.031.2%
Hard jobs done in a real computer terminal: building and fixing software, setting up systems, and science and engineering problems.
16th of 17. The maker's own number, from DeepSeek, via LLM Stats.
- Humanity's Last Exam36.8%
About 2,500 questions written by experts in more than a hundred fields to be as hard as they could make them, many needing a diagram or image.
32nd of 74. The maker's own number, from DeepSeek, via LLM Stats.
- CursorBench47.7%
Ambiguous coding tasks that span many files, taken from real sessions in the Cursor code editor.
9th of 20. Independent result from Epoch AI.
- DeepSWE 1.174.2%
Software engineering tasks run in the same simple agent setup for every model.
7th of 27. The maker's own number, from DeepSeek, via LLM Stats.
- SimpleBench66.7%
Multiple-choice puzzles about everyday situations, space, time and people, written to trip up models that pattern-match instead of think. Ordinary adults do well on them.
28th of 68. Independent result from Epoch AI.
- SimpleQA Verified42.9%
Short factual questions with one checked answer, asked without web search.
45th of 78. Independent result from Epoch AI.
- FrontierMath71.2%
Hundreds of new, unpublished math problems written by mathematicians, from hard undergraduate level up to research, with answers a computer can check.
30th of 76. Independent result from Epoch AI.
- LMArena text1,463
People put a question to two anonymous models and vote for the better answer.
30th of 107. Independent result from LMArena.
- LMArena WebDev1,664
People ask two anonymous models to build the same web page or app and vote for the better one.
15th of 98. Independent result from LMArena.
6 more results in its full profile.
Every result with its source, what it costs per task, what jumped with its release, side by side with any other model, and answers to your questions about when to use it.
Models ranked near it
How the rating works, and every benchmark it uses: the model leaderboard.