DeepSeek-V3 (Mar 2025)
#137 of 274 scored.
DeepSeekMar 24, 2025Open weights
Capabilities Index
Range 132.1–137.8
135.9
Context
Tokens
—
Input
Per 1M tokens
—
Output
Per 1M tokens
—
Benchmarks
How it scores.
- GPQA Diamond56.8%
104th of 186
- Mock AIME37.7%
117th of 176
Everything else Epoch has for it
- Lech Mazur Writing77.0%
- MATH level 575.5%
- Aider polyglot55.1%
- Fiction.LiveBench50.0%
- METR Time Horizons49.6%
- DTBench41.3%
- WeirdML36.1%
- LMCA18.2%
- SimpleBench12.6%
On the index
Around it.
Sources