Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Claude Sonnet 4.6
84
Winner · 3/8 categoriesSarvam 105B
60
0/8 categoriesClaude Sonnet 4.6· Sarvam 105B
Pick Claude Sonnet 4.6 if you want the stronger benchmark profile. Sarvam 105B only becomes the better choice if you want the cheaper token bill or you want the stronger reasoning-first profile.
Claude Sonnet 4.6 is clearly ahead on the aggregate, 84 to 60. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Claude Sonnet 4.6's sharpest advantage is in agentic, where it averages 72.5 against 49.5. The single biggest benchmark swing on the page is SWE-bench Verified, 79.6% to 45%.
Claude Sonnet 4.6 is also the more expensive model on tokens at $3.00 input / $15.00 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for Sarvam 105B. That is roughly Infinityx on output cost alone. Sarvam 105B is the reasoning model in the pair, while Claude Sonnet 4.6 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. Claude Sonnet 4.6 gives you the larger context window at 200K, compared with 128K for Sarvam 105B.
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 | Claude Sonnet 4.6 | Sarvam 105B |
|---|---|---|
| AgenticClaude Sonnet 4.6 wins | ||
| OSWorld-Verified | 72.5% | — |
| Claw-Eval | 66.3% | — |
| BrowseComp | — | 49.5% |
| CodingClaude Sonnet 4.6 wins | ||
| SWE-bench Verified | 79.6% | 45% |
| FLTEval | 23.7% | — |
| SWE-Rebench | 60.7% | — |
| React Native Evals | 77.9% | — |
| LiveCodeBench v6 | — | 71.7% |
| Multimodal & Grounded | ||
| MMMU-Pro | 95% | — |
| OfficeQA Pro | 88% | — |
| Reasoning | ||
| MuSR | 93% | — |
| BBH | 88% | — |
| LongBench v2 | 83% | — |
| gpqaDiamond | — | 78.7% |
| hle | — | 11.2% |
| Knowledge | ||
| MMLU | 99% | 90.6% |
| MMLU-Pro | — | 81.7% |
| Instruction Following | ||
| IFEval | — | 84.8% |
| Multilingual | ||
| MGSM | 91% | — |
| MMLU-ProX | 89% | — |
| MathematicsClaude Sonnet 4.6 wins | ||
| AIME 2025 | 98% | 88.3% |
| HMMT Feb 2025 | 96% | — |
| BRUMO 2025 | 96% | — |
| MATH-500 | — | 98.6% |
| HMMT Feb 2025 | — | 85.8% |
| HMMT Nov 2025 | — | 85.8% |
Claude Sonnet 4.6 is ahead overall, 84 to 60. The biggest single separator in this matchup is SWE-bench Verified, where the scores are 79.6% and 45%.
Claude Sonnet 4.6 has the edge for coding in this comparison, averaging 66.4 versus 45. Inside this category, SWE-bench Verified is the benchmark that creates the most daylight between them.
Claude Sonnet 4.6 has the edge for math in this comparison, averaging 97.1 versus 92.3. Inside this category, AIME 2025 is the benchmark that creates the most daylight between them.
Claude Sonnet 4.6 has the edge for agentic tasks in this comparison, averaging 72.5 versus 49.5. Sarvam 105B stays close enough that the answer can still flip depending on your workload.
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.