- Terminal Bench 2.1 83.9
- NL2Repo 57.7
- Cybergym 75.3
- DeepSWE 59.3
- Toolathlon-Verified 75.9
- DSBench-Hard (in-house) 63.6
- AutomationBench (Public) 25.7
- Agents Last Exam 27.3
- Chartography 64.3
- ZeroBench (Pass@5) 35
DeepSeek-V4-Flash-Vision-Exp
DeepSeek-V4-Flash-Vision-Exp places in the top three on 2 benchmarks.
16
Benchmarks with a score
0
Ranked first, out of five or more models
2
In the top three
DeepSeek Aug 2026 DeepSeek-V4-Flash-Vision-Exp
Open weights Mid-size
157.5
#47 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- NL2Repo-Bench 57.7% #9 /30
- ZeroBench 35% #14 /76
- Chartography 11.2% #39 /43
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Hemingway-bench | 981 | #34 of 42 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| NL2Repo-Bench | 57.7% | #9 of 30 | Lab model card | |
| DeepSWE | 59.3% | #42 of 56 | Lab model card | |
| Vision2Web | 67.6% | #8 of 10 | Official leaderboard | |
| FrontierSWE | 14.8% | #23 of 25 | Official leaderboard | |
| DSBench (DeepSeek internal) | 63.6% | #3 of 3 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Terminal-Bench 2.1 | 83.9% | #20 of 56 | Lab model card | |
| Toolathlon-Verified | 75.9% | #8 of 21 | Lab model card | |
| AutomationBench | 25.7% | #38 of 47 | Lab model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Agents' Last Exam | 27.3% | #17 of 26 | Lab model card | |
| Tuesday Work Index | 34.3 | #37 of 44 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ZeroBench | 35% | #14 of 76 | Official leaderboard | |
| Chartography | 11.2% | #39 of 43 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AILuminate Safety 1.2 | 38.0 | #3 of 5 | Official leaderboard | |
| MLCommons AILuminate | 38% | #3 of 5 | Official leaderboard | |
| CyberGym | 75.3% | #22 of 34 | Lab model card |
Its lab's numbers put it at 157.5, number 47 of 404
The Self-Reported Index fits one score to the 8 capability results DeepSeek reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 154.2 to 160.8. How we compute it.
- Terminal-Bench 2.1 83.9
- Toolathlon-Verified 75.9
- Chartography 64.3
- DeepSWE 59.3
- NL2Repo-Bench 57.7
- ZeroBench 35
- Agents' Last Exam 27.3
- AutomationBench 25.7
Its lab's numbers next to 2 leaderboards
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.