Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Claude Opus 4.6
85
Winner · 3/8 categoriesGemma 4 31B
73
1/8 categoriesClaude Opus 4.6· Gemma 4 31B
Pick Claude Opus 4.6 if you want the stronger benchmark profile. Gemma 4 31B only becomes the better choice if coding is the priority or you want the cheaper token bill.
Claude Opus 4.6 is clearly ahead on the aggregate, 85 to 73. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Claude Opus 4.6's sharpest advantage is in knowledge, where it averages 77.8 against 61.3. The single biggest benchmark swing on the page is HLE, 53% to 26.5%. Gemma 4 31B does hit back in coding, so the answer changes if that is the part of the workload you care about most.
Claude Opus 4.6 is also the more expensive model on tokens at $15.00 input / $75.00 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for Gemma 4 31B. That is roughly Infinityx on output cost alone. Gemma 4 31B is the reasoning model in the pair, while Claude Opus 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 Opus 4.6 gives you the larger context window at 1M, compared with 256K for Gemma 4 31B.
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 Opus 4.6 | Gemma 4 31B |
|---|---|---|
| Agentic | ||
| Terminal-Bench 2.0 | 65.4% | — |
| BrowseComp | 84% | — |
| OSWorld-Verified | 72.7% | — |
| BrowseComp-VL | 35.9% | — |
| OSWorld | 72.2% | — |
| Tau2-Airline | 82.0% | — |
| Tau2-Telecom | 92.1% | — |
| PinchBench | 93.3% | — |
| BFCL v4 | 77.0% | — |
| AndroidWorld | 62.0% | — |
| WebVoyager | 88.0% | — |
| Claw-Eval | 66.3% | — |
| CodingGemma 4 31B wins | ||
| HumanEval | 91% | — |
| SWE-bench Verified | 80.8% | — |
| SWE-bench Verified* | 75.6% | — |
| LiveCodeBench | 76% | 80% |
| FLTEval | 39.6% | — |
| SWE-Rebench | 65.3% | — |
| React Native Evals | 84.4% | — |
| Multimodal & GroundedClaude Opus 4.6 wins | ||
| MMMU-Pro | 77.3% | 76.9% |
| OfficeQA Pro | 94% | — |
| Design2Code | 77.3% | — |
| Flame-VLM-Code | 98.8% | — |
| Vision2Web | 43.5% | — |
| MMSearch | 63.8% | — |
| MMSearch-Plus | 25.6% | — |
| SimpleVQA | 63.2% | — |
| V* | 66.5% | — |
| ReasoningClaude Opus 4.6 wins | ||
| MuSR | 93% | — |
| BBH | 94% | 74.4% |
| LongBench v2 | 92% | — |
| MRCRv2 | 76% | 66.4% |
| ARC-AGI-2 | 68.8% | — |
| KnowledgeClaude Opus 4.6 wins | ||
| MMLU | 99% | — |
| GPQA | 91.3% | 84.3% |
| GPQA-D | 89.2% | — |
| SuperGPQA | 95% | — |
| MMLU-Pro | 82% | 85.2% |
| MMLU-Pro (Arcee) | 89.1% | — |
| HLE | 53% | 26.5% |
| FrontierScience | 88% | — |
| SimpleQA | 72% | — |
| HLE w/o tools | — | 19.5% |
| Instruction Following | ||
| IFEval | 95% | — |
| IFBench | 53.1% | — |
| Multilingual | ||
| MGSM | 96% | — |
| MMLU-ProX | 94% | — |
| Mathematics | ||
| AIME 2024 | 99% | — |
| AIME 2025 | 98% | — |
| AIME25 (Arcee) | 99.8% | — |
| HMMT Feb 2025 | 96% | — |
| BRUMO 2025 | 96% | — |
| MATH-500 | 98% | — |
Claude Opus 4.6 is ahead overall, 85 to 73. The biggest single separator in this matchup is HLE, where the scores are 53% and 26.5%.
Claude Opus 4.6 has the edge for knowledge tasks in this comparison, averaging 77.8 versus 61.3. Inside this category, HLE is the benchmark that creates the most daylight between them.
Gemma 4 31B has the edge for coding in this comparison, averaging 80 versus 72. Inside this category, LiveCodeBench is the benchmark that creates the most daylight between them.
Claude Opus 4.6 has the edge for reasoning in this comparison, averaging 82.4 versus 66.4. Inside this category, BBH is the benchmark that creates the most daylight between them.
Claude Opus 4.6 has the edge for multimodal and grounded tasks in this comparison, averaging 84.8 versus 76.9. Inside this category, MMMU-Pro 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.