Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Gemma 4 26B A4B
64
Winner · 1/8 categoriesGrok 3 Mini
~49
1/8 categoriesGemma 4 26B A4B· Grok 3 Mini
Pick Gemma 4 26B A4B if you want the stronger benchmark profile. Grok 3 Mini only becomes the better choice if knowledge is the priority.
Gemma 4 26B A4B is clearly ahead on the aggregate, 64 to 49. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Gemma 4 26B A4B's sharpest advantage is in coding, where it averages 77.1 against 41.5. The single biggest benchmark swing on the page is LiveCodeBench, 77.1% to 41.5%. Grok 3 Mini does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.
Grok 3 Mini is also the more expensive model on tokens at $0.30 input / $0.50 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for Gemma 4 26B A4B. That is roughly Infinityx on output cost alone. Gemma 4 26B A4B gives you the larger context window at 256K, compared with 128K for Grok 3 Mini.
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 | Gemma 4 26B A4B | Grok 3 Mini |
|---|---|---|
| Agentic | ||
| Coming soon | ||
| CodingGemma 4 26B A4B wins | ||
| LiveCodeBench | 77.1% | 41.5% |
| Multimodal & Grounded | ||
| MMMU-Pro | 73.8% | — |
| Reasoning | ||
| BBH | 64.8% | — |
| MRCRv2 | 44.1% | — |
| KnowledgeGrok 3 Mini wins | ||
| GPQA | 82.3% | 66.2% |
| MMLU-Pro | 82.6% | 78.9% |
| HLE | 17.2% | — |
| HLE w/o tools | 8.7% | — |
| SimpleQA | — | 21.7% |
| Instruction Following | ||
| Coming soon | ||
| Multilingual | ||
| Coming soon | ||
| Mathematics | ||
| AIME 2024 | — | 95.8% |
Gemma 4 26B A4B is ahead overall, 64 to 49. The biggest single separator in this matchup is LiveCodeBench, where the scores are 77.1% and 41.5%.
Grok 3 Mini has the edge for knowledge tasks in this comparison, averaging 59.8 versus 56.1. Inside this category, GPQA is the benchmark that creates the most daylight between them.
Gemma 4 26B A4B has the edge for coding in this comparison, averaging 77.1 versus 41.5. Inside this category, LiveCodeBench 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.