1 vs 0 benchmarks won
DeepSeek DeepSeek-V4-Pro | Meta Muse Spark | |
|---|---|---|
| Overview | ||
| Company | DeepSeek | Meta |
| Release date | Apr 24 2026 | Apr 8 2026 |
| Model type | — | — |
| Open source | Yes | No |
| Specifications | ||
Parameters | — | — |
Context window | — | — |
| Benchmarks | ||
Coding SWE-Bench VerifiedReal coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. | — | 77.4% |
Science GPQA DiamondGraduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 90.1%Best | 89.5% |
Multimodal MMMUTests the AI on understanding images and text together across many college subjects. Higher is better. | — | 80.4% |
| Timeline | ||
| Release gap | Muse Spark shipped 16 days before DeepSeek-V4-Pro | |
DeepSeek-V4-Pro leads Muse Spark on 1 of the 1 benchmark they both report (GPQA Diamond). Muse Spark shipped 16 days before DeepSeek-V4-Pro, so benchmark comparisons should account for the intervening progress.
DeepSeek-V4-Pro is an open-source / open-weight model; Muse Spark is proprietary.
On GPQA Diamond, DeepSeek-V4-Pro leads at 90.1% vs Muse Spark at 89.5%.