BlogLLM Benchmark Scores as Published, Side by Side
Gemini 3.1 Pro Preview: Benchmark Scores as Published
Every result the catalog holds for Gemini 3.1 Pro Preview: 147 tests in 17 areas, from 19 pages. 54 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 2026-02-19Context window 1.05MMax output 65.5KInput Text, image, video, audio, PDFOutput TextWeights ClosedData version 2026103001, as of 2026-10-30
Knowledge Work
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SWE-bench Verified | 80.6% | Google, Gemini 3.1 Pro model cardchecked 2026-10-15; printed on 5 more pages | Effort: high |
| SWE-Bench Pro | 54.2% | Google, Gemini 3.1 Pro model cardchecked 2026-10-15; printed on 14 more pages | Effort: high |
| Terminal-Bench 2.1 | 70.3% | Google, Gemini 3.5 Flash model cardchecked 2026-10-15; printed on 7 more pages | Harness: Terminus-2 |
| DeepSWE v1.1 | 12% | Google, Gemini 3.6 Flash model cardchecked 2026-10-15; printed on 2 more pages | - |
| SWE-bench Multilingual | 76.9% | Moonshot AI, Kimi K2.6 model cardchecked 2026-10-15 | Effort: high; Harness: in-house SWE-agent variant |
| Terminal-Bench 2.0 | 68.5% | Google, Gemini 3.1 Pro model cardchecked 2026-10-15; printed on 5 more pages | Effort: high; Harness: Terminus-2 |
| NL2Repo-Bench | 21.62 | MiniMax, MiniMax M3checked 2026-10-15; printed on 2 more pages | - |
| SWE-Marathon | 4% | Z.ai, GLM-5.2 model cardchecked 2026-10-15 | Context: 1000000 |
| ProgramBench | 39.5% | Z.ai, GLM-5.2 model cardchecked 2026-10-15 | Harness: Claude Code 2.1.156; Context: 400000 |
| SWE-Atlas-QnA (setting not stated) | 13.5% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| FrontierSWE (dominance score) | 39.6 | Z.ai, GLM-5.2 model cardchecked 2026-10-15 | Context: 1000000 |
| Artificial Analysis Coding Agent Index v1.1 | 42.7 | OpenAI, GPT-5.6: Frontier intelligence that scales with your ambitionchecked 2026-10-15 | - |
| SWE-Atlas-Test Writing | 29.84% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| SWE-fficiency | 19.7% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| VIBE-Pro | 41% | MiniMax, MiniMax-M2.7 model cardchecked 2026-10-15 | - |
| VIBE-V2 (MiniMax internal) | 28% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
ML Engineering
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| PostTrainBench v1.1 | 21.6% | Z.ai, GLM-5.2 model cardchecked 2026-10-15 | Context: 1000000 |
| PaperBench | 46.7 | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| MLE-bench | 42.6% | Google, Gemini 3.6 Flash model cardchecked 2026-10-15 | - |
| MLE-bench Lite (medal rate) | 66.6% | MiniMax, MiniMax-M2.7 model cardchecked 2026-10-15 | - |
| KernelBench Hard (TFLOPs relative to hardware peak) | 18.6% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| PostTrainBench (version not stated) | 15.2% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
Science and Math
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MRCR v2 8-needle 128K | 84.9% | Google, Gemini 3.1 Pro model cardchecked 2026-10-15; printed on 2 more pages | Effort: high; Context: 128k (average) |
| MRCR v2 8-needle 1M (pointwise) | 26.3% | Google, Gemini 3.1 Pro model cardchecked 2026-10-15; printed on 2 more pages | Effort: high; Context: 1M (pointwise) |
| CorpusQA 1M | 53.8% | DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15 | Effort: high; Context: 1M |
| MRCR 1M (MMR) | 76.3% | DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15 | Effort: high; Context: 1M |
Computer Use
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| OSWorld-Verified | 76.2% | Google, Gemini 3.5 Flash model cardchecked 2026-10-15; printed on 4 more pages | - |
Multimodal Understanding
Cybersecurity
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| CyberGym | 38.8% | Z.ai, GLM-5.1 model cardchecked 2026-10-15 | - |
General Knowledge and Reasoning
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveCodeBench v6 | 91.7% | Moonshot AI, Kimi K2.6 model cardchecked 2026-10-15 | Effort: high |
| Codeforces rating | 3052 | DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15 | Effort: high |
| LiveCodeBench (version not stated) | 91.7% | DeepSeek, DeepSeek-V4-Pro (preview) model cardchecked 2026-10-15 | Effort: high |
| LiveSQLBench (Base-Full v1) | 39.83% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| SVG-Bench (MiniMax internal) | 59.2% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| OJBench (Python) | 70.7% | Moonshot AI, Kimi K2.6 model cardchecked 2026-10-15 | Effort: high |
| LiveCodeBench Pro (Elo) | 2887 | Google, Gemini 3.1 Pro model cardchecked 2026-10-15 | Effort: high |
Tool Use and Workflows
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| BrowseComp | 85.9% | Google, Gemini 3.1 Pro model cardchecked 2026-10-15; printed on 8 more pages | Effort: high; Tools: search, python, browse |
| DeepSearchQA (accuracy) | 60.2% | Moonshot AI, Kimi K2.6 model cardchecked 2026-10-15 | Effort: high; Tools: search, code interpreter, browsing |
| DeepSearchQA (F1) | 81.9% | Moonshot AI, Kimi K2.6 model cardchecked 2026-10-15 | Effort: high; Tools: search, code interpreter, browsing |
Instruction Following
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| IFBench | 77.1 | Thinking Machines, Thinking Machines Inkling model cardchecked 2026-10-15 | Effort: high |
Multilingual
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMMLU | 92.6% | Google, Gemini 3.1 Pro model cardchecked 2026-10-15; printed on 1 more page | Effort: high |
| Global-MMLU-Lite | 92.7% | Thinking Machines, Thinking Machines Inkling model cardchecked 2026-10-15 | Effort: high |
Safety and Alignment
Engineering and Design
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Blueprint-Bench 2 (normalized score) | 26.5% | Google, Gemini 3.5 Flash model cardchecked 2026-10-15; printed on 1 more page | - |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Artificial Analysis Intelligence Index | 57 | MiniMax, MiniMax-M2.7 model cardchecked 2026-10-15 | - |
| Artificial Analysis Intelligence Index v4.1 | 46.5 | OpenAI, GPT-5.6: Frontier intelligence that scales with your ambitionchecked 2026-10-15 | - |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked. Cost per task is measured by Vals AI.
| Price | Value | Printed by | Setup as printed |
|---|---|---|---|
| Input | $2 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $4 per 1M tokens above 200,000 tokens |
| Cached input | $0.2 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $0.4 per 1M tokens above 200,000 tokens |
| Output | $12 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $18 per 1M tokens above 200,000 tokens |
| Input, batch | $1 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $2 per 1M tokens above 200,000 tokens |
| Cached input, batch | $0.2 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $0.4 per 1M tokens above 200,000 tokens |
| Output, batch | $6 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $9 per 1M tokens above 200,000 tokens |
| Input, flex | $1 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $2 per 1M tokens above 200,000 tokens |
| Cached input, flex | $0.2 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $0.4 per 1M tokens above 200,000 tokens |
| Output, flex | $6 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $9 per 1M tokens above 200,000 tokens |
| Input, priority | $3.6 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $7.2 per 1M tokens above 200,000 tokens |
| Cached input, priority | $0.36 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $0.72 per 1M tokens above 200,000 tokens |
| Output, priority | $21.6 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $32.4 per 1M tokens above 200,000 tokens |
| Input, contributor | $1 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $2 per 1M tokens above 200,000 tokens |
| Cached input, contributor | $0.1 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $0.2 per 1M tokens above 200,000 tokens |
| Output, contributor | $6 | Google, Gemini API pricingchecked 2026-10-15 | Context: up to 200,000 tokens; Long context: $9 per 1M tokens above 200,000 tokens |
| Cost per taskVals AI, USD, lower is better | $3.83 | Vals AI, Vibe Code Bench v1.1 leaderboardchecked 2026-10-27 | Harness: OpenHands |
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.