GoogleGemini 3.1 Flash Lite
Gemini 3.1 Flash Lite is a Google proprietary reasoning model, released 3 Mar 2026. Its best configuration ranks #274 of 785 on the BenchLeader Index at 53.7 ±4.0, in the upper half. The ± is the point: 287 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 long context (62) and lowest in agents & tools (43). At $0.563 per million tokens blended it is mid-priced. Output speed of 390 tokens per second puts it in the fastest quarter, with a first answer in 4.5 s. It has been measured at 4 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.
- Blended price
- $0.563/M
- $0.250 in · $1.50 out
- Output speed
- 390 tok/s
- measured by Artificial Analysis
- First answer
- 4.50 s
- first token 0.82 s
- Context
- 1.0M
- Full answer
- 5.79 s
- median, reasoning included
- Cost per run
- $0.060
- one full Intelligence Index run
- Released
- 3 Mar 2026
- Overall index54
- Reasoning47
- Coding47
- Agents & tools43
- Maths61
- Knowledge53
- Instruction following61
- Human preference61
- Multimodal58
- Long context62
- Composite47
Versions
Google has shipped 6 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 3.5 Flash Lite | 21 Jul 2026 | 54.5 | #252 |
| Gemini 3.1 Flash Litethis page | 3 Mar 2026 | 53.7 | #274 |
| Gemini 2.5 Flash Lite 09thinking | 25 Sept 2025 | 47.9 | #444 |
| Gemini 2.5 Flash-Lite | 17 Jun 2025 | 48.3 | #433 |
| Gemini 2.0 Flash Lite 001 | 25 Feb 2025 | – | – |
| Gemini 2.0 Flash-Lite | 5 Feb 2025 | 47.2 | #462 |
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 |
|---|---|---|---|---|---|---|
| minimal | – | – | 390 tok/s | 4.50 s | $0.0006 | Maths 38 · Reasoning 54 |
| low | – | – | 390 tok/s | 4.50 s | $0.0006 | Maths 39 · Reasoning 54 |
| high | 48.8 | #415 | 390 tok/s | 4.50 s | $0.0006 | Agents & tools 30 · Knowledge 55 · Maths 50 · Reasoning 59 |
| not statedbest | 53.7 | #274 | 390 tok/s | 4.50 s | $0.0006 | Agents & tools 43 · Coding 47 · Composite 47 · Human preference 61 · Instruction following 61 · Knowledge 53 · Long context 62 · Maths 61 · Multimodal 58 · Reasoning 47 |
How its index has moved
8 Sept 2026 to 21 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
See all LLM benchmarks. 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 | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| GPQA Diamond | 73.7%#157 | 74.2%#154 | 81.8%#117 | – | Epoch AI Benchmarking Hub | |
| Humanity's Last Exameffort not stated | – | – | – | 8.6%#31 | Scale AI / CAIS | |
| Arena Hard Promptseffort not stated | – | – | – | 1445#116 | Arena (LMArena) | |
| GPQA Diamond (AA)not in indexeffort not stated | – | – | – | 82.2%#173 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in indexeffort not stated | – | – | – | 17.2%#237 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | 81.1%#63 | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | 67.2%#34 | Kagi LLM Benchmark | |
| CritPteffort not stated | – | – | – | 1.1%#223 | Artificial Analysis | |
| LMCAnot in index | – | – | 35.0%#125 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | 76.8%#144 | – | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | – | 62.7%#51 | ForecastBench |
Coding
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| SciCodeeffort not stated | – | 43.4%#144 | SciCode | |
| WeirdMLeffort not stated | – | 52.2%#68 | WeirdML | |
| Arena Codingeffort not stated | – | 1458#136 | Arena (LMArena) | |
| Arena WebDeveffort not stated | – | 1256#123 | Arena (LMArena) | |
| SciCode (AA)not in indexeffort not stated | – | 43.4%#154 | Artificial Analysis | |
| LiveCodeBencheffort not stated | – | 80.1%#68 | Vals AI | |
| SWE-bench (Vals)not in index | 62.8%#72 | – | Vals AI | |
| Code Migrationnot in index | 4.6%#70 | – | Vals AI | |
| ProgramBenchnot in index | 0.0%#32 | – | Vals AI | |
| Vibe Code Bench v1.1not in index | 0.0%#111 | – | Vals AI | |
| ALE-Benchnot in indexeffort not stated | – | 797.7#74 | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in indexeffort not stated | – | 0.5%#139 | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in indexeffort not stated | – | 31.1%#152 | Artificial Analysis |
Agents & tools
Maths
| Benchmark | minimal | low | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 21.4%#86 | 22.5%#84 | 27.7%#77 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 37.8%#200 | 44.4%#193 | 80.0%#114 | – | Epoch AI Benchmarking Hub | |
| AIME (Vals) | – | – | 83.3%#43 | – | Vals AI | |
| Arena Mathseffort not stated | – | – | – | 1436#97 | Arena (LMArena) |
Knowledge
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| AA-Omniscienceeffort not stated | – | -16.4#227 | Artificial Analysis | |
| MMLU-Proeffort not stated | – | 86.2%#43 | Vals AI | |
| LegalBench | 83.8%#48 | – | Vals AI | |
| CorpFin | 59.4%#76 | – | Vals AI | |
| TaxEval | 71.8%#70 | – | Vals AI | |
| MultiNRCeffort not statedpossible contamination | – | 25.0%#29 | Scale AI SEAL | |
| CaseLaw v2not in index | 55.0%#35 | – | Vals AI | |
| Excel Modeling Benchmarknot in index | 8.6%#73 | – | Vals AI | |
| MedCodenot in index | 47.6%#32 | – | Vals AI | |
| MedScribenot in index | 63.9%#98 | – | Vals AI | |
| AA-Omniscience: accuracynot in indexeffort not stated | – | 36.3%#151 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in indexeffort not stated | – | 17.3%#316 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 77.2%#17 | Artificial Analysis | |
| MultiChallenge | 60.6%#14 | Scale AI SEAL | |
| TutorBench | 51.5%#15 | Scale AI SEAL | |
| Arena Instruction Followingnot in index | 1409#125 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1433#107 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1414#87 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1435#106 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1429#122 | Arena (LMArena) |
Multimodal
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Visioneffort not stated | – | 1240#72 | Arena (LMArena) | |
| MMMU-Proeffort not stated | – | 75.5%#97 | Artificial Analysis | |
| VISTAeffort not stated | – | 46.9%#19 | Scale AI SEAL | |
| MMMU-Pro (Vals)not in index | 82.5%#37 | – | Vals AI | |
| MortgageTaxnot in index | 68.0%#20 | – | Vals AI | |
| SAGEnot in index | 49.5%#20 | – | Vals AI |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 74.3%#162 | Artificial Analysis | |
| MLCRnot in index | 5.0%#144 | Artificial Analysis |
Composite
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | 144.3#88 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2effort not stated | – | 15.6#280 | Artificial Analysis | |
| Vals Multimodal Indexnot in index | 41.4%#29 | – | Vals AI |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 51.0%#12 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 48.4%#57 | Scale AI SEAL |
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 | 206 tok/s | 0.48 s | $0.125 | $0.750 | 1.0M | – |
| Google AI Studio | 185 tok/s | 0.73 s | $0.250 | $1.50 | 1.0M | – |
| Google Vertex | 115 tok/s | 0.81 s | $0.250 | $1.50 | 1.0M | – |
| Google Vertex | 115 tok/s | 1.03 s | $0.450 | $2.70 | 1.0M | – |
| Google AI Studio | 111 tok/s | 0.59 s | $0.450 | $2.70 | 1.0M | – |
| Google Vertex | 98 tok/s | 0.60 s | $0.275 | $1.65 | 1.0M | – |
| Google Vertex | 62 tok/s | 0.97 s | $0.275 | $1.65 | 1.0M | – |
| Google Vertex | 30 tok/s | 17 s | $0.125 | $0.750 | 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 $0.025 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0006 | $0.0005 | 5.3 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0039 | $0.0019 | 6.0 s |
| Code edit | 6,000 / 1,500 | $0.0038 | $0.0027 | 8.4 s |
| Agentic coding session | 60,000 / 4,000 | $0.021 | $0.011 | 14.8 s |
| Structured extraction | 2,000 / 200 | $0.0008 | $0.0005 | 5.0 s |
More from Google
See also
Questions people ask
- What are Gemini 3.1 Flash Lite's benchmark scores?
- As of 24 Oct 2026, Gemini 3.1 Flash Lite scores 53.7 ±4.0 on the BenchLeader Index, strongest in long context (62) and weakest in agents & tools (43). That rests on 24 independent benchmark results, and every score on this page links to its publisher.
- How does Gemini 3.1 Flash Lite rank?
- Gemini 3.1 Flash Lite is #274 of 785 ranked configurations on 24 Oct 2026. 287 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Gemini 3.1 Flash Lite cost per 1M tokens?
- As of 24 Oct 2026, Gemini 3.1 Flash Lite costs $0.250 per 1M input tokens and $1.50 per 1M output tokens, $0.563 blended (three input tokens per output token).
- When was Gemini 3.1 Flash Lite released?
- Google released Gemini 3.1 Flash Lite on 3 Mar 2026. BenchLeader has results for it at 4 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Gemini 3.1 Flash Lite: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gemini-3-1-flash-lite.