DeepSeek-V4-Pro
#61 of 274 scored.
DeepSeekApr 24, 2026Open weights
Capabilities Index
Range 147.3–150.7
149.1
Context
Takes text
1.05M
Input
Per 1M tokens
$0.21
Output
Per 1M tokens
$0.42
Benchmarks
How it scores.
- GPQA Diamond87.9%
30th of 186
- SWE-bench Verified77.6%
6th of 32
- FrontierMath45.3%
49th of 81
- SimpleQA Verified47.0%
33rd of 77
- Mock AIME96.7%
28th of 176
Everything else Epoch has for it
- DTBench84.5%
- WeirdML48.9%
- LMCA48.5%
- Surface Evolver Bench40.0%
- FrontierCode17.6%
- ProofBench16.0%
- Chess Puzzles15.8%
- CL-bench Life13.5%
- Mystery Game Puzzles8.6%
- FrontierMath-Tier-4-v2-Private2.4%
On the index
Around it.
Sources