BlogLLM Benchmark Scores as Published, Side by Side
Gemma 4 31B: Benchmark Scores as Published
Every result the catalog holds for Gemma 4 31B: 58 tests in 10 areas, from 3 pages. 15 printed by Google itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer GoogleReleased 2026-04-02Context window 256KInput Text, imageOutput TextWeights Open (Apache 2.0)Data version 2026103001, as of 2026-10-30
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SWE-bench Verified | 52% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | Harness: internal scaffold; Context: 200000 |
| SWE-Bench Pro | 35.7% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | Harness: internal scaffold; Context: 200000 |
| SWE-bench Multilingual | 51.7% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | Harness: internal scaffold; Context: 200000 |
| Terminal-Bench 2.0 | 42.9% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | Harness: Harbor/Terminus-2; Context: 256000 |
| NL2Repo-Bench | 15.5 | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | Harness: Claude Code (baselines) |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 84.3% | Google, Gemma 4 model cardchecked 2026-10-15; printed on 2 more pages | - |
| HMMT February 2025 | 88.7% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| IMO-AnswerBench | 74.5% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| AIME 2026 | 89.2% | Google, Gemma 4 model cardchecked 2026-10-15; printed on 2 more pages | Tools: none |
| HMMT November 2025 | 87.5% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| HMMT February 2026 | 77.2% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MRCR v2 8-needle 128K | 66.4% | Google, Gemma 4 model cardchecked 2026-10-15 | Context: 128k (average) |
Multimodal Understanding
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Humanity's Last Exam (no tools) | 19.5% | Google, Gemma 4 model cardchecked 2026-10-15; printed on 2 more pages | Tools: none |
| MMLU-Pro | 85.2% | Google, Gemma 4 model cardchecked 2026-10-15; printed on 2 more pages | - |
| Humanity's Last Exam (with tools) | 26.5% | Google, Gemma 4 model cardchecked 2026-10-15 | Tools: search |
| MMLU-Redux | 93.7% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| C-Eval | 82.6% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| SuperGPQA | 65.7% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| BIG-Bench Extra Hard | 74.4% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveCodeBench v6 | 80% | Google, Gemma 4 model cardchecked 2026-10-15; printed on 2 more pages | - |
| Codeforces rating | 2150 | Google, Gemma 4 model cardchecked 2026-10-15 | - |
| QwenWebBench rating | 1197 | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
Tool Use and Workflows
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Toolathlon | 21.2% | Alibaba, Qwen3.6-35B-A3B model cardchecked 2026-10-15 | - |
| MCP Atlas (public set) | 57.2% | Alibaba, Qwen3.6-35B-A3B model cardchecked 2026-10-15 | - |
| MCPMark | 18.1% | Alibaba, Qwen3.6-35B-A3B model cardchecked 2026-10-15 | - |
| Claw-Eval pass^3 | 25% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepPlanning | 24% | Alibaba, Qwen3.6-35B-A3B model cardchecked 2026-10-15 | - |
| VitaBench | 43% | Alibaba, Qwen3.6-35B-A3B model cardchecked 2026-10-15 | - |
| τ³-bench (domain not stated) | 67.5% | Alibaba, Qwen3.6-35B-A3B model cardchecked 2026-10-15 | - |
| SkillsBench | 23.6 | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | Harness: OpenCode |
| τ²-bench (average of 3 domains) | 76.9% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
| Claw-Eval average score | 48.5 | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| QwenClawBench | 41.7 | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | Context: 256000 |
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| WideSearch (metric not stated) | 35.2 | Alibaba, Qwen3.6-35B-A3B model cardchecked 2026-10-15 | - |
Multilingual
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMMLU | 88.4% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
Healthcare
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MedXpertQA MM | 61.3% | Google, Gemma 4 model cardchecked 2026-10-15 | - |
More from Google
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.