1 vs 0 benchmarks won
Anthropic Claude 3 Haiku | DeepSeek DeepSeek-MoE | |
|---|---|---|
| Overview | ||
| Company | Anthropic | DeepSeek |
| Release date | Mar 4 2024 | Jan 9 2024 |
| Model type | — | — |
| Open source | No | Yes |
| Specifications | ||
Parameters | — | — |
Context window | — | — |
| Benchmarks | ||
Science reasoning GPQA Diamond | 33.3% | — |
Software engineering SWE-Bench Verified | — | — |
Multimodal understanding MMMU | — | — |
| Timeline | ||
| Release gap | DeepSeek-MoE shipped 55 days before Claude 3 Haiku | |
Claude 3 Haiku leads DeepSeek-MoE on 1 of the tracked benchmarks (GPQA Diamond, SWE-Bench Verified, MMMU). DeepSeek-MoE shipped 55 days before Claude 3 Haiku, so benchmark comparisons should account for the intervening progress.
DeepSeek-MoE is an open-source / open-weight model; Claude 3 Haiku is proprietary.
Direct benchmark comparisons are unavailable — at least one of these models has not published scores on GPQA Diamond, SWE-Bench Verified, or MMMU.