DeepSeek-V2 (MoE-236B, May 2024)
#184 of 274 scored.
DeepSeekMay 7, 2024Open weights
Capabilities Index
Range 118.5–127.8
124.8
Context
Tokens
—
Input
Per 1M tokens
—
Output
Per 1M tokens
—
Benchmarks
No independent scores yet.
Epoch AI hasn't run DeepSeek-V2 (MoE-236B, May 2024) yet. Its scores show up here the morning after they do.
Everything else Epoch has for it
- ARC AI289.6%
- HellaSwag82.8%
- TriviaQA80.0%
- Winogrande72.6%
- BBH71.7%
- MMLU71.2%
- PIQA67.8%
On the index
Around it.
Sources