GoogleGemini 2.5 Flash
Gemini 2.5 Flash is a Google proprietary reasoning model, released 17 Apr 2025. Its best configuration ranks #306 of 785 on the BenchLeader Index at 52.7 ±3.9, in the upper half. The ± is the point: 289 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 (63) and lowest in reasoning (41). At $0.850 per million tokens blended it is mid-priced. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.
- Blended price
- $0.850/M
- $0.300 in · $2.50 out
- Output speed
- –
- First answer
- –
- Context
- 1.0M
- Full answer
- –
- Cost per run
- –
- Released
- 17 Apr 2025
- Overall index53
- Reasoning41
- Coding44
- Agents & tools48
- Maths55
- Knowledge48
- Instruction following63
- Human preference59
- Multimodal62
- Long context58
Versions
Google has shipped 12 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 3.8 Flashhigh | 2 Sept 2026 | 64.4 | #50 |
| Gemini 3.7 Flashhigh | 13 Aug 2026 | 63.5 | #64 |
| Gemini 3.6 Flashhigh | 21 Jul 2026 | 62.0 | #89 |
| Gemini 3.5 Flashhigh | 19 May 2026 | 63.1 | #70 |
| Gemini 3 Flashthinking | 17 Dec 2025 | 59.1 | #138 |
| Gemini 2.5 Flash 09 2025thinking | 25 Sept 2025 | 51.1 | #351 |
| Gemini 2.5 Flashthis page | 17 Apr 2025 | 52.7 | #306 |
| Gemini 2.0 Flash | 5 Feb 2025 | 43.2 | #574 |
| Gemini 1.5 Flash 002 | 24 Sept 2024 | 38.3 | #716 |
| Gemini 1.5 Flash 001 | 23 May 2024 | – | – |
| Gemini 1.5 Flash | 14 May 2024 | 44.3 | #541 |
| Gemini Flash 2.0 | – | – | – |
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 |
|---|---|---|---|---|---|---|
| no reasoning | 46.0 | #501 | 206 tok/s | 0.36 s | $0.0009 | Agents & tools 44 · Composite 40 · Instruction following 42 · Knowledge 42 · Long context 49 · Multimodal 48 · Reasoning 43 |
| thinking | 47.7 | #448 | 224 tok/s | 17 s | $0.0009 | Agents & tools 46 · Coding 36 · Composite 44 · Instruction following 52 · Knowledge 51 · Long context 57 · Multimodal 52 · Reasoning 40 |
| not statedbest | 52.7 | #306 | – | – | $0.0009 | Agents & tools 48 · Coding 44 · Human preference 59 · Instruction following 63 · Knowledge 48 · Long context 58 · Maths 55 · Multimodal 62 · Reasoning 41 |
How its index has moved
8 Sept 2026 to 23 Oct 2026- not stated
- thinking
- no reasoning
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 | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Humanity's Last Exameffort not stated | – | – | 12.1%#28 | Scale AI / CAIS | |
| SimpleBencheffort not stated | – | – | 41.2%#58 | SimpleBench | |
| Arena Hard Promptseffort not stated | – | – | 1420#147 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 68.3%#314 | 79.0%#203 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 4.7%#477 | 12.1%#281 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in indexeffort not stated | – | – | 57.6%#109 | Vals AI | |
| Kagi LLM Benchmark | – | 56.8%#60 | 44.1%#97 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | – | 39.0%#50 | HELM Capabilities | |
| ARC-AGI-1 | – | 33.3%#202 | 33.3%#202 | ARC Prize | |
| ARC-AGI-2 | – | 2.5%#199 | 1.7%#210 | ARC Prize | |
| CritPt | 1.4%#211 | 1.1%#223 | – | Artificial Analysis | |
| LMCAnot in indexeffort not stated | – | – | 27.5%#145 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | – | 76.5%#145 | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | 62.9%#50 | ForecastBench |
Coding
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| WeirdML | 41.0%#99 | 41.0%#99 | WeirdML | |
| Arena Codingeffort not stated | – | 1427#176 | Arena (LMArena) | |
| LiveCodeBench | 46.9%#114 | 56.9%#105 | Vals AI | |
| Aider Polygloteffort not stated | – | 55.1%#19 | Aider polyglot leaderboard | |
| SWE-bench Verified (bash only)effort not stated | – | 28.7%#38 | SWE-bench | |
| SWE-bench Verified (any scaffold)not in indexeffort not stated | – | 28.7%#52 | SWE-bench | |
| ALE-Benchnot in indexeffort not stated | – | 661.9#90 | Epoch AI Benchmarking Hub |
Agents & tools
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Terminal-Bencheffort not stated | – | – | 17.1%#76 | Terminal-Bench | |
| Terminal-Bench Hard | 12.1%#210 | 13.6%#201 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 14.9%#363 | 31.6%#251 | – | Artificial Analysis | |
| BFCL Overalleffort not stated | – | – | 56.2%#13 | Berkeley Function Calling Leaderboard | |
| BALROGnot in indexeffort not stated | – | – | 33.5%#13 | BALROG | |
| Vending-Bench 2not in indexeffort not stated | – | – | 548.8#58 | Andon Labs | |
| BrowseComp-Plusnot in indexeffort not stated | – | – | 16.3%#11 | BrowseComp-Plus | |
| BrowseCompnot in indexeffort not stated | – | – | 2.5%#9 | OpenAI | |
| DeepSearchQAnot in indexeffort not stated | – | – | 25.9%#8 |
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIMEeffort not stated | – | 73.1%#134 | Epoch AI Benchmarking Hub | |
| Omni-MATH (HELM)effort not stated | – | 38.4%#32 | HELM Capabilities | |
| MATH 500not in index | 91.8%#19 | 91.6%#21 | Vals AI | |
| Arena Mathseffort not stated | – | 1409#136 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| AA-Omniscience | -42.6#336 | -29.8#279 | – | Artificial Analysis | |
| LegalBencheffort not stated | – | – | 83.8%#47 | Vals AI | |
| CorpFineffort not stated | – | – | 54.2%#92 | Vals AI | |
| TaxEval | – | 70.5%#87 | 71.2%#79 | Vals AI | |
| MedQA | – | 91.0%#42 | 86.7%#56 | Vals AI | |
| PRBench Financeeffort not stated | – | – | 38.4%#31 | Scale AI SEAL | |
| PRBench Legaleffort not stated | – | – | 41.0%#28 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | – | 63.9%#43 | HELM Capabilities | |
| MedCodenot in index | – | 40.4%#63 | 38.4%#69 | Vals AI | |
| MedScribenot in index | – | 83.0%#46 | 82.9%#48 | Vals AI | |
| AA-Omniscience: accuracynot in index | 26.1%#219 | 25.9%#222 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 7.0%#493 | 24.7%#245 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 39.0%#274 | 50.3%#173 | – | Artificial Analysis | |
| IFEval (HELM)effort not stated | – | – | 89.8%#10 | HELM Capabilities | |
| WildBench (HELM)effort not stated | – | – | 81.7%#25 | HELM Capabilities | |
| Arena Instruction Followingnot in indexeffort not stated | – | – | 1400#138 | Arena (LMArena) |
Human preference
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Texteffort not stated | – | 1409#141 | Arena (LMArena) | |
| Arena Creative Writingnot in indexeffort not stated | – | 1394#113 | Arena (LMArena) | |
| Arena Multi-turnnot in indexeffort not stated | – | 1403#150 | Arena (LMArena) | |
| Arena Longer Queriesnot in indexeffort not stated | – | 1417#138 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 76.5%#22 | – | Lech Mazur |
Multimodal
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Arena Visioneffort not stated | – | – | 1254#61 | Arena (LMArena) | |
| MMMU-Pro | 65.5%#174 | 69.1%#158 | – | Artificial Analysis | |
| VISTAeffort not stated | – | – | 49.1%#13 | Scale AI SEAL | |
| MMMU (validation)effort not stated | – | – | 79.7%#6 | MMMU | |
| MMMU-Pro (Vals)not in index | – | 71.9%#64 | 68.0%#71 | Vals AI | |
| MortgageTaxnot in index | – | 57.5%#71 | 58.4%#68 | Vals AI | |
| SAGEnot in index | – | 39.6%#59 | 44.8%#45 | Vals AI | |
| AudioMCnot in index | – | 40.0%#9 | 26.1%#15 | Scale AI SEAL | |
| AudioMC (text output)not in index | – | 40.0%#7 | 26.1%#10 | Scale AI SEAL | |
| VTBnot in indexeffort not stated | – | – | 4.7%#16 | Scale AI SEAL | |
| GeoBenchnot in indexeffort not stated | – | – | 76.0%#7 | GeoBench | |
| VPCTnot in indexeffort not stated | – | – | 38.0%#21 | Epoch AI Benchmarking Hub | |
| SpatialViz-Benchnot in indexeffort not stated | – | – | 36.9%#3 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Fiction.LiveBench 120keffort not stated | – | – | 68.8%#8 | Fiction.live | |
| AA-LCR | 49.9%#324 | 65.3%#250 | – | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | 142.9#97 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 9.8#389 | 13.1#318 | – | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | – | 62.6%#39 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| MASKnot in indexpossible contamination | 49.1%#53 | Scale AI SEAL | |
| PropensityBenchnot in index | 68.0%#4 | Scale AI SEAL |
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 $0.030 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0009 | $0.0008 | – |
| Summarize a 30-page report | 12,000 / 600 | $0.0051 | $0.0027 | – |
| Code edit | 6,000 / 1,500 | $0.0056 | $0.0043 | – |
| Agentic coding session | 60,000 / 4,000 | $0.028 | $0.016 | – |
| Structured extraction | 2,000 / 200 | $0.0011 | $0.0007 | – |
More from Google
See also
Questions people ask
- What are Gemini 2.5 Flash's benchmark scores?
- As of 24 Oct 2026, Gemini 2.5 Flash scores 52.7 ±3.9 on the BenchLeader Index, strongest in instruction following (63) and weakest in reasoning (41). That rests on 29 independent benchmark results, and every score on this page links to its publisher.
- How does Gemini 2.5 Flash rank?
- Gemini 2.5 Flash is #306 of 785 ranked configurations on 24 Oct 2026. 289 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 2.5 Flash cost per 1M tokens?
- As of 24 Oct 2026, Gemini 2.5 Flash costs $0.300 per 1M input tokens and $2.50 per 1M output tokens, $0.850 blended (three input tokens per output token).
- When was Gemini 2.5 Flash released?
- Google released Gemini 2.5 Flash on 17 Apr 2025. BenchLeader has results for it at 3 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Gemini 2.5 Flash: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gemini-2-5-flash.