GoogleReasoning modelGemini 3.1 Pro
Gemini 3.1 Pro is a Google proprietary reasoning model, released 5 Mar 2026. Its best configuration ranks #78 of 756 on the BenchLeader Index at 62.6 ±4.1, in the upper half. The ± is the point: 157 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (74) and lowest in agents & tools (57). At $4.50 per million tokens blended it is among the most expensive ranked models. Output speed of 117 tokens per second puts it faster than most, with a first answer in 25.1 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 2 Oct 2026.
- Blended price
- $4.50/M
- $2.00 in · $12.00 out
- Output speed
- 117 tok/s
- measured by Artificial Analysis
- First answer
- 25 s
- first token 3.04 s
- Context
- 1.0M
- Full answer
- 29 s
- median, reasoning included
- Cost per run
- $0.675
- one full Intelligence Index run
- Released
- 5 Mar 2026
- Overall index63
- Reasoning69
- Coding57
- Agents & tools57
- Maths61
- Knowledge65
- Instruction following74
- Human preference59
- Multimodal65
- Long context66
- Composite64
Versions
Google has shipped 11 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Gemini 3.1 Prothis page | 5 Mar 2026 | 62.6 | #78 |
| Gemini 3 Pro | 18 Nov 2025 | 60.8 | #101 |
| Gemini 2.5 Pro | 25 Mar 2025 | 54.0 | #247 |
| Gemini 1.5 Pro | 15 Feb 2025 | 42.5 | #573 |
| Gemini 2.0 Pro | 5 Feb 2025 | 47.1 | #445 |
| Gemini 1.5 Pro 002 | 24 Sept 2024 | 44.7 | #507 |
| Gemini 1.5 Pro 001 | 14 May 2024 | 44.7 | #508 |
| Gemini 1.0 Pro 001 | 13 Dec 2023 | – | – |
| Gemini Pro | – | – | – |
| Gemini 1.0 Pro | – | – | – |
| Gemini 1.0 Pro 002 | – | – | – |
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| low | – | – | 117 tok/s | 25 s | $0.0044 | |
| medium | – | – | 117 tok/s | 25 s | $0.0044 | |
| high | 59.5 | #121 | 117 tok/s | 25 s | $0.0044 | Agents & tools 58 · Coding 58 · Composite 55 · Knowledge 66 · Maths 66 · Reasoning 65 |
| thinking | – | – | 117 tok/s | 25 s | $0.0044 | Coding 61 |
| not statedbest | 62.6 | #78 | 117 tok/s | 25 s | $0.0044 | Agents & tools 57 · Coding 57 · Composite 64 · Human preference 59 · Instruction following 74 · Knowledge 65 · Long context 66 · Maths 61 · Multimodal 65 · Reasoning 69 |
How its index has moved
8 Sept 2026 to 5 Oct 2026- not stated
- high
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
Coding
Agents & tools
Maths
| Benchmark | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | – | – | – | 59.6%#34 | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | – | 26.8%#38 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | – | – | 95.6%#36 | – | 95.6%#35 | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | – | 26.0%#43 | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | 91.0%#29 | – | – | LiveBench | |
| AIME (Vals) | – | – | 98.1%#1 | – | – | Vals AI | |
| AIME 2026 | – | – | – | – | 98.3%#4 | MathArena | |
| HMMT February 2026 | – | – | – | – | 94.7%#8 | MathArena | |
| MathArena Apex | – | – | – | – | 60.9%#5 | MathArena | |
| LMArena Maths | – | – | – | – | 1488#28 | LMArena |
Knowledge
Instruction following
| Benchmark | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | 85.4%#14 | – | – | LiveBench | |
| LiveBench Instruction Followingnot in index | – | – | 79.1%#3 | – | – | LiveBench | |
| IFBench | – | – | – | – | 77.1%#18 | Artificial Analysis | |
| MultiChallenge | – | – | – | – | 71.4%#3 | Scale AI SEAL | |
| LMArena Instruction Followingnot in index | – | – | – | – | 1480#20 | LMArena |
Human preference
| Benchmark | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| LMArena Text | – | – | – | – | 1487#18 | LMArena | |
| EQ-Bench 4 | – | – | – | – | 1142#20 | EQ-Bench | |
| LMArena Creative Writingnot in index | – | – | – | – | 1480#11 | LMArena | |
| LMArena Multi-turnnot in index | – | – | – | – | 1497#12 | LMArena | |
| LMArena Longer Queriesnot in index | – | – | – | – | 1499#17 | LMArena |
Multimodal
| Benchmark | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| LMArena Vision | – | – | – | – | 1296#21 | LMArena | |
| MMMU-Pro | – | – | – | – | 82.4%#32 | Artificial Analysis | |
| MMMU-Pro (official)not in index | – | – | 80.5%#2 | – | – | MMMU | |
| MMMU-Pro (Vals)not in index | – | – | 88.2%#10 | – | – | Vals AI | |
| MortgageTaxnot in index | – | – | 69.4%#6 | – | – | Vals AI | |
| SAGEnot in index | – | – | 48.7%#26 | – | – | Vals AI | |
| VTBnot in index | – | – | – | – | 29.0%#3 | Scale AI SEAL | |
| Blueprint-Bench 2not in index | – | – | – | – | 26.5%#20 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| AA-LCR | – | – | – | – | 82.0%#41 | Artificial Analysis | |
| LMArena Documentnot in index | – | – | – | – | 1459#27 | LMArena |
Composite
| Benchmark | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | – | – | 154.8#30 | Epoch AI Benchmarking Hub | |
| LiveBench | – | – | 77.0%#25 | – | – | LiveBench | |
| AA Intelligence Index | – | – | – | – | 29.7#112 | Artificial Analysis | |
| Vals Indexnot in index | – | – | 33.4#35 | – | – | Vals AI | |
| Vals Multimodal Indexnot in index | – | – | 56.1%#20 | – | – | Vals AI |
Safety & honesty
| Benchmark | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| FORTRESSnot in index | – | – | – | – | 29.8%#25 | Scale AI SEAL | |
| MASKnot in index | – | – | – | – | 42.4%#64 | Scale AI SEAL |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Google AI Studio | 100 tok/s | 3.18 s | $1.00 | $6.00 | 1.0M | – |
| Google Vertex | 94 tok/s | 3.17 s | $2.00 | $12.00 | 1.0M | – |
| Google AI Studio | 92 tok/s | 3.41 s | $2.00 | $12.00 | 1.0M | – |
| Google Vertex | 18 tok/s | 95 s | $1.00 | $6.00 | 1.0M | – |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.200 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0044 | $0.0039 | 27.7 s |
| Summarise a 30-page report | 12,000 / 600 | $0.031 | $0.015 | 30.3 s |
| Code edit | 6,000 / 1,500 | $0.030 | $0.022 | 37.9 s |
| Agentic coding session | 60,000 / 4,000 | $0.168 | $0.087 | 59.2 s |
| Structured extraction | 2,000 / 200 | $0.0064 | $0.0037 | 26.9 s |
See also
Data as of 5 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Gemini 3.1 Pro: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/gemini-3-1-pro, data as of 5 Oct 2026.