DeepSeek-V3
DeepSeek·
LLMsopen-weightcloud + localFrontier
Overview
Latest iteration with improved benchmarks.
Capabilities and innovations
SOTA Open WeightsFP8 TrainingMulti-token predictionFP8 TrainingMulti-token PredictionAuxiliary Loss-Free Load Balancing
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 59.1% | unsourced |
| Humanity’s Last Exam (no tools) | 11.8% | unsourced |
| SWE-bench Verified | 42% | unsourced |
| MMLU | 88.5% | unsourced |
- SWE-bench Pro: benchmark did not exist at release
- MMLU-Pro: benchmark did not exist at release
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
Architecture and hardware
- Parameters
- 671B total, 37B active per token (MoE)
- Estimated VRAM at Q4
- ~386 GB, Frontier class
Links
More from DeepSeek
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.