Rankings / Fireworks AI
Deepseek V3p2
release date unknown
BenchAtlas Index
as of 2026-07-14
Benchmark evidence
14 results
knowledge science
| GPQA Diamond implementation=vals-ai | high effort | 80.3 % | independent | Vals AI |
| GPQA Diamond implementation=vals-ai | no reasoning | 76.3 % | independent | Vals AI |
| MMLU-Pro implementation=vals-ai | high effort | 84.9 % | independent | Vals AI |
| MMLU-Pro implementation=vals-ai | no reasoning | 83.1 % | independent | Vals AI |
professional
| CaseLaw | high effort | 55.4 % | independent | Vals AI |
| CorpFin | high effort | 51.0 % | independent | Vals AI |
| CorpFin | no reasoning | 47.9 % | independent | Vals AI |
| LegalBench | high effort | 76.1 % | independent | Vals AI |
| LegalBench | no reasoning | 66.0 % | independent | Vals AI |
| MedQA | high effort | 93.9 % | independent | Vals AI |
| MedQA | no reasoning | 89.5 % | independent | Vals AI |
| TaxEval | high effort | 68.2 % | independent | Vals AI |
reasoning math
| AIME implementation=vals-ai | high effort | 84.6 % | independent | Vals AI |
| AIME implementation=vals-ai | no reasoning | 64.8 % | independent | Vals AI |
