Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
DeepSeekMath V2
61
Winner · 3/8 categoriesGranite-4.0-350M
~27
0/8 categoriesDeepSeekMath V2· Granite-4.0-350M
Pick DeepSeekMath V2 if you want the stronger benchmark profile. Granite-4.0-350M only becomes the better choice if you would rather avoid the extra latency and token burn of a reasoning model.
DeepSeekMath V2 is clearly ahead on the aggregate, 61 to 27. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
DeepSeekMath V2's sharpest advantage is in multilingual, where it averages 82.5 against 16.2. The single biggest benchmark swing on the page is MGSM, 87% to 16.2%.
DeepSeekMath V2 is the reasoning model in the pair, while Granite-4.0-350M is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. DeepSeekMath V2 gives you the larger context window at 128K, compared with 32K for Granite-4.0-350M.
BenchLM keeps the benchmark table and the operator tradeoffs on the same page so a better score does not hide a materially slower, pricier, or smaller-context model.
Runtime metrics show N/A when BenchLM does not have a sourced snapshot for that exact model. The scoring rules and freshness policy are documented on the methodology page.
| Benchmark | DeepSeekMath V2 | Granite-4.0-350M |
|---|---|---|
| Agentic | ||
| Terminal-Bench 2.0 | 65% | — |
| BrowseComp | 66% | — |
| OSWorld-Verified | 61% | — |
| Coding | ||
| HumanEval | 72% | 38% |
| SWE-bench Verified | 45% | — |
| LiveCodeBench | 44% | — |
| SWE-bench Pro | 51% | — |
| Multimodal & Grounded | ||
| MMMU-Pro | 64% | — |
| OfficeQA Pro | 73% | — |
| Reasoning | ||
| MuSR | 75% | — |
| BBH | 86% | 33.3% |
| LongBench v2 | 75% | — |
| MRCRv2 | 72% | — |
| KnowledgeDeepSeekMath V2 wins | ||
| MMLU | 80% | 36.2% |
| GPQA | 79% | 26.1% |
| SuperGPQA | 77% | — |
| MMLU-Pro | 74% | 14.4% |
| HLE | 18% | — |
| FrontierScience | 73% | — |
| SimpleQA | 77% | — |
| Instruction FollowingDeepSeekMath V2 wins | ||
| IFEval | 83% | 61.6% |
| MultilingualDeepSeekMath V2 wins | ||
| MGSM | 87% | 16.2% |
| MMLU-ProX | 80% | — |
| Mathematics | ||
| AIME 2023 | 80% | — |
| AIME 2024 | 82% | — |
| AIME 2025 | 81% | — |
| HMMT Feb 2023 | 76% | — |
| HMMT Feb 2024 | 78% | — |
| HMMT Feb 2025 | 77% | — |
| BRUMO 2025 | 79% | — |
| MATH-500 | 90% | — |
DeepSeekMath V2 is ahead overall, 61 to 27. The biggest single separator in this matchup is MGSM, where the scores are 87% and 16.2%.
DeepSeekMath V2 has the edge for knowledge tasks in this comparison, averaging 62.3 versus 18.5. Inside this category, MMLU-Pro is the benchmark that creates the most daylight between them.
DeepSeekMath V2 has the edge for instruction following in this comparison, averaging 83 versus 61.6. Inside this category, IFEval is the benchmark that creates the most daylight between them.
DeepSeekMath V2 has the edge for multilingual tasks in this comparison, averaging 82.5 versus 16.2. Inside this category, MGSM is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.