GoogleGemini 3.8 Flash
Best configuration ranks #45 of 610 on the BenchLeader Index at 63.5 ±5.2 (medium reasoning effort). Last measured 2 Sept 2026. Released 2 Sept 2026.
- Blended price
- $1.50/M
- $0.750 in · $3.75 out
- Output speed
- 273 tok/s
- First answer
- 14 s
- first token 1.59 s
- Context
- 1M
- Overall index64
- Coding64
- Knowledge78
- Multimodal69
- Long context69
- Composite79
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 |
|---|---|---|---|---|---|---|
| low | 61.9 | #61 | 273 tok/s | 14 s | $0.0014 | Coding 63 · Composite 72 · Knowledge 74 · Long context 67 · Multimodal 69 |
| mediumbest | 63.5 | #45 | 273 tok/s | 14 s | $0.0014 | Coding 64 · Composite 79 · Knowledge 78 · Long context 69 · Multimodal 69 |
| high | 60.1 | #93 | 273 tok/s | 14 s | $0.0014 | Agents & tools 48 · Coding 65 · Composite 54 · Human preference 70 · Knowledge 67 · Maths 60 · Reasoning 69 |
| default | 62.4 | #55 | 273 tok/s | 14 s | $0.0014 | Agents & tools 57 · Composite 81 · Knowledge 70 · Long context 67 · Maths 57 · Multimodal 70 |
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 | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| GPQA Diamond | – | – | 95.4%#2 | – | Epoch AI Benchmarking Hub | |
| LMArena Hard Prompts | – | – | 1518#9 | – | LMArena | |
| LiveBench Reasoningnot in index | – | – | 89.3%#12 | – | LiveBench | |
| GPQA Diamond (AA)not in index | 92.0%#41 | 93.5%#14 | – | 95.3%#3 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 37.1%#72 | 42.1%#44 | – | 47.8%#20 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | 94.4%#4 | – | Vals AI |
Coding
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| SciCode | 54.3%#31 | 54.4%#30 | 53.6%#37 | – | SciCode | |
| LMArena Coding | – | – | 1537#7 | – | LMArena | |
| LMArena WebDev | – | – | 1568#20 | – | LMArena | |
| LiveBench Codingnot in index | – | – | 72.5%#43 | – | LiveBench | |
| SciCode (AA)not in index | 55.0%#34 | 55.1%#33 | – | 56.6%#19 | Artificial Analysis | |
| LiveCodeBench | – | – | 89.5%#3 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | – | 80.0%#24 | – | Vals AI |
Agents & tools
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| Terminal-Bench | – | – | 19.1%#56 | – | Terminal-Bench | |
| LMArena Agent | – | – | 4#13 | – | LMArena | |
| LiveBench Agentic Codingnot in index | – | – | 54.2%#24 | – | LiveBench | |
| Terminal-Bench 2.1 (Vals) | – | – | 81.3%#5 | – | Vals AI | |
| HiL-Bench | – | – | – | 41.5%#6 | Scale AI SEAL |
Maths
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | – | 68.4%#19 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | 21.9%#38 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | – | – | 98.9%#11 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | 48.0%#21 | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | 91.6%#17 | – | LiveBench |
Knowledge
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | 69.7%#5 | – | Epoch AI Benchmarking Hub | |
| LiveBench Data Analysisnot in index | – | – | 54.0%#51 | – | LiveBench | |
| AA-Omniscience | 21.4#37 | 28.6#21 | – | 29.6#18 | Artificial Analysis | |
| MMLU-Pro | – | – | 90.2%#5 | – | Vals AI | |
| LegalBench | – | – | 87.0%#6 | – | Vals AI | |
| TaxEval | – | – | 74.5%#35 | – | Vals AI | |
| PRBench Finance | – | – | – | 49.5%#9 | Scale AI SEAL | |
| PRBench Legal | – | – | – | 48.9%#11 | Scale AI SEAL |
Instruction following
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | 87.8%#5 | – | LiveBench |
Human preference
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| LMArena Text | – | – | 1494#8 | – | LMArena |
Multimodal
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| MMMU-Pro | 84.5%#11 | 84.2%#13 | – | 85.6%#4 | Artificial Analysis |
Long context
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| AA-LCR | 80.7%#41 | 84.0%#6 | – | 81.3%#34 | Artificial Analysis |
Composite
| Benchmark | low | medium | high | default | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench | – | – | 75.8%#25 | – | LiveBench | |
| AA Intelligence Index | 33.8#64 | 40#35 | – | 41.2#31 | Artificial Analysis | |
| Vals Indexnot in index | – | – | 62.3#7 | – | 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 | 258 tok/s | 1.54 s | $1.35 | $6.75 | 1.0M | – |
| Google AI Studio | 157 tok/s | 1.49 s | $0.750 | $3.75 | 1.0M | – |
| Google AI Studio Flex | 151 tok/s | 1.65 s | $0.375 | $1.88 | 1.0M | – |
| Google Vertex Priority | 87 tok/s | 1.38 s | $1.35 | $6.75 | 1.0M | – |
| Google Vertex | 60 tok/s | 2.31 s | $0.750 | $3.75 | 1.0M | – |
| Google Vertex Flex | 37 tok/s | 13 s | $0.375 | $1.88 | 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 $0.075 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0014 | $0.0012 | 14.9 s |
| Summarise a 30-page report | 12,000 / 600 | $0.011 | $0.0052 | 16.0 s |
| Code edit | 6,000 / 1,500 | $0.010 | $0.0071 | 19.3 s |
| Agentic coding session | 60,000 / 4,000 | $0.060 | $0.030 | 28.5 s |
| Structured extraction | 2,000 / 200 | $0.0023 | $0.0012 | 14.6 s |
See also
Data as of 9 Sept 2026. Compare these configurations.