GoogleReasoning modelGemini 3 Flash
Best configuration ranks #66 of 610 on the BenchLeader Index at 61.5 ±3.3 (thinking reasoning effort). Released 17 Dec 2025.
- Blended price
- $1.13/M
- $0.500 in · $3.00 out
- Output speed
- 200 tok/s
- First answer
- 6.12 s
- first token 0.85 s
- Context
- 1M
- Overall index62
- Agents & tools67
- Knowledge69
- Instruction following75
- Multimodal65
- Long context66
- Composite62
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. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| minimal | 53.4 | #226 | 202 tok/s | 0.85 s | $0.0011 | Coding 51 · Human preference 65 · Multimodal 61 · Reasoning 42 |
| low | – | – | 202 tok/s | 0.85 s | $0.0011 | Reasoning 35 |
| medium | – | – | 202 tok/s | 0.85 s | $0.0011 | Reasoning 43 |
| high | 58.5 | #121 | 202 tok/s | 0.85 s | $0.0011 | Agents & tools 52 · Coding 63 · Knowledge 64 · Maths 63 · Reasoning 58 |
| thinkingbest | 61.5 | #66 | 200 tok/s | 6.12 s | $0.0011 | Agents & tools 67 · Composite 62 · Instruction following 75 · Knowledge 69 · Long context 66 · Multimodal 65 |
| default | 56.6 | #153 | 202 tok/s | 0.85 s | $0.0011 | Agents & tools 53 · Coding 53 · Composite 52 · Human preference 67 · Instruction following 55 · Knowledge 62 · Long context 54 · Maths 53 · Multimodal 64 · Reasoning 61 |
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
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| GPQA Diamond | – | – | – | 89.4%#43 | – | 83.2%#97 | Epoch AI Benchmarking Hub | |
| SimpleBench | – | – | – | – | – | 61.1%#26 | SimpleBench | |
| LMArena Hard Prompts | 1475#62 | – | – | – | – | 1493#35 | LMArena | |
| GPQA Diamond (AA)not in index | – | – | – | – | 89.8%#68 | 81.2%#177 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | – | – | – | – | 36.6%#74 | 15.0%#194 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | 87.9%#36 | – | – | Vals AI | |
| ARC-AGI-1 | 21.5%#154 | 29.0%#146 | 57.7%#112 | 84.7%#76 | – | – | ARC Prize | |
| ARC-AGI-2 | 3.3%#134 | 1.3%#153 | 12.8%#102 | 33.6%#83 | – | – | ARC Prize |
Coding
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SWE-bench Verified (Epoch) | – | – | – | – | – | 75.4%#13 | Epoch AI Benchmarking Hub | |
| WeirdML | – | – | – | – | – | 61.6%#42 | WeirdML | |
| GSO-Bench | – | – | – | – | – | 9.8%#17 | GSO-Bench | |
| LMArena Coding | 1491#76 | – | – | – | – | 1507#52 | LMArena | |
| LMArena WebDev | 1383#80 | – | – | – | – | 1438#58 | LMArena | |
| LiveCodeBench | – | – | – | 85.6%#29 | – | – | Vals AI | |
| IOI | – | – | – | 39.1%#13 | – | – | Vals AI | |
| SWE-bench (Vals)not in index | – | – | – | 75.0%#42 | – | – | Vals AI | |
| SWE-Bench Pro | – | – | – | – | – | 34.6%#15 | Scale AI SEAL | |
| SWE-bench Verified (bash only) | – | – | – | 75.8%#2 | – | – | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | – | – | – | 75.8%#5 | – | – | SWE-bench |
Agents & tools
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Terminal-Bench | – | – | – | – | – | 64.3%#11 | Terminal-Bench | |
| APEX-Agents | – | – | – | – | – | 24.0%#31 | Mercor | |
| Terminal-Bench Hard | – | – | – | – | 38.6%#59 | 31.8%#99 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | – | – | – | – | 80.4%#119 | 43.3%#210 | Artificial Analysis | |
| Terminal-Bench 2.1 (Vals) | – | – | – | 53.9%#42 | – | – | Vals AI | |
| MCP Atlas | – | – | – | – | – | 62.0%#23 | Scale AI SEAL | |
| τ²-bench | – | – | – | 83.5%#4 | – | – | τ²-bench |
Maths
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | – | – | – | – | 51.2%#38 | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | – | – | 17.1%#42 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | – | – | – | 95.6%#29 | – | 92.8%#50 | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | – | – | 15.0%#44 | Vals AI | |
| AIME (Vals) | – | – | – | 95.6%#8 | – | – | Vals AI | |
| MGSM | – | – | – | 93.3%#10 | – | – | Vals AI | |
| AIME 2026 | – | – | – | – | – | 96.7%#6 | MathArena | |
| HMMT February 2026 | – | – | – | – | – | 89.4%#14 | MathArena | |
| MathArena Apex | – | – | – | – | – | 15.6%#16 | MathArena |
Knowledge
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | – | 66.8%#8 | – | – | Epoch AI Benchmarking Hub | |
| AA-Omniscience | – | – | – | – | 10.1#64 | -4.3#113 | Artificial Analysis | |
| MMLU-Pro | – | – | – | 88.6%#19 | – | – | Vals AI | |
| LegalBench | – | – | – | 86.9%#9 | – | – | Vals AI | |
| CorpFin | – | – | – | 66.4%#19 | – | – | Vals AI | |
| TaxEval | – | – | – | 73.9%#44 | – | – | Vals AI | |
| MedQA | – | – | – | 95.8%#11 | – | – | Vals AI |
Instruction following
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| IFBench | – | – | – | – | 78.0%#14 | 55.1%#145 | Artificial Analysis |
Human preference
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Text | 1459#53 | – | – | – | – | 1474#30 | LMArena |
Multimodal
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Vision | 1266#37 | – | – | – | – | 1285#24 | LMArena | |
| MMMU-Pro | – | – | – | – | 79.9%#37 | 78.5%#48 | Artificial Analysis |
Long context
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| AA-LCR | – | – | – | – | 78.0%#82 | 55.3%#233 | Artificial Analysis |
Composite
| Benchmark | minimal | low | medium | high | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | – | – | – | 151.8#38 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index | – | – | – | – | 26.3#108 | 17.9#195 | Artificial Analysis | |
| Vals Indexnot in index | – | – | – | 29.4#42 | – | – | Vals AI |
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 Priority | 84 tok/s | 1.25 s | $0.900 | $5.40 | 1.0M | – |
| Google AI Studio | 83 tok/s | 1.34 s | $0.500 | $3.00 | 1.0M | – |
| Google Vertex Priority | 78 tok/s | 1.63 s | $0.900 | $5.40 | 1.0M | – |
| Google Vertex | 58 tok/s | 1.19 s | $0.500 | $3.00 | 1.0M | – |
| Google Vertex Flex | 50 tok/s | 19 s | $0.250 | $1.50 | 1.0M | – |
| Google AI Studio Flex | 7 tok/s | 1.94 s | $0.250 | $1.50 | 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 $0.050 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0011 | $0.0010 | 7.6 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0078 | $0.0038 | 9.1 s |
| Code edit | 6,000 / 1,500 | $0.0075 | $0.0055 | 13.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.042 | $0.022 | 26.1 s |
| Structured extraction | 2,000 / 200 | $0.0016 | $0.0009 | 7.1 s |
See also
Data as of 9 Sept 2026. Compare these configurations.