Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
DeepSeek V3.2 (Thinking)
67
Winner · 3/8 categoriesSarvam 105B
60
2/8 categoriesDeepSeek V3.2 (Thinking)· Sarvam 105B
Pick DeepSeek V3.2 (Thinking) if you want the stronger benchmark profile. Sarvam 105B only becomes the better choice if knowledge is the priority.
DeepSeek V3.2 (Thinking) is clearly ahead on the aggregate, 67 to 60. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
DeepSeek V3.2 (Thinking)'s sharpest advantage is in agentic, where it averages 69.4 against 49.5. The single biggest benchmark swing on the page is BrowseComp, 70% to 49.5%. Sarvam 105B does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.
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 | DeepSeek V3.2 (Thinking) | Sarvam 105B |
|---|---|---|
| AgenticDeepSeek V3.2 (Thinking) wins | ||
| Terminal-Bench 2.0 | 71% | — |
| BrowseComp | 70% | 49.5% |
| OSWorld-Verified | 67% | — |
| DeepPlanning | 27.4% | — |
| CodingDeepSeek V3.2 (Thinking) wins | ||
| HumanEval | 79% | — |
| SWE-bench Verified | 48% | 45% |
| LiveCodeBench | 45% | — |
| SWE-bench Pro | 58% | — |
| LiveCodeBench v6 | — | 71.7% |
| Multimodal & Grounded | ||
| MMMU-Pro | 66% | — |
| Reasoning | ||
| MuSR | 81% | — |
| BBH | 86% | — |
| LongBench v2 | 78% | — |
| MRCRv2 | 78% | — |
| ARC-AGI-2 | 4% | — |
| gpqaDiamond | — | 78.7% |
| hle | — | 11.2% |
| KnowledgeSarvam 105B wins | ||
| MMLU | 87% | 90.6% |
| GPQA | 85% | — |
| SuperGPQA | 83% | — |
| MMLU-Pro | 73% | 81.7% |
| HLE | 22% | — |
| FrontierScience | 77% | — |
| SimpleQA | 83% | — |
| Instruction FollowingDeepSeek V3.2 (Thinking) wins | ||
| IFEval | 85% | 84.8% |
| Multilingual | ||
| MGSM | 84% | — |
| MMLU-ProX | 79% | — |
| MathematicsSarvam 105B wins | ||
| AIME 2023 | 87% | — |
| AIME 2024 | 89% | — |
| AIME 2025 | 88% | 88.3% |
| HMMT Feb 2023 | 83% | — |
| HMMT Feb 2024 | 85% | — |
| HMMT Feb 2025 | 84% | — |
| BRUMO 2025 | 86% | — |
| MATH-500 | 84% | 98.6% |
| HMMT Feb 2025 | — | 85.8% |
| HMMT Nov 2025 | — | 85.8% |
DeepSeek V3.2 (Thinking) is ahead overall, 67 to 60. The biggest single separator in this matchup is BrowseComp, where the scores are 70% and 49.5%.
Sarvam 105B has the edge for knowledge tasks in this comparison, averaging 81.7 versus 65.9. Inside this category, MMLU-Pro is the benchmark that creates the most daylight between them.
DeepSeek V3.2 (Thinking) has the edge for coding in this comparison, averaging 50.7 versus 45. Inside this category, SWE-bench Verified is the benchmark that creates the most daylight between them.
Sarvam 105B has the edge for math in this comparison, averaging 92.3 versus 86.3. Inside this category, MATH-500 is the benchmark that creates the most daylight between them.
DeepSeek V3.2 (Thinking) has the edge for agentic tasks in this comparison, averaging 69.4 versus 49.5. Inside this category, BrowseComp is the benchmark that creates the most daylight between them.
DeepSeek V3.2 (Thinking) has the edge for instruction following in this comparison, averaging 85 versus 84.8. Inside this category, IFEval 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.