GoogleGemini 2.5 Pro
Gemini 2.5 Pro is a Google proprietary reasoning model, released 25 Mar 2025. Its best configuration ranks #264 of 785 on the BenchLeader Index at 53.9 ±3.2, in the upper half. The ± is the point: 279 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 (66) and lowest in agents & tools (41). At $3.44 per million tokens blended it is among the most expensive ranked models. Output speed of 121 tokens per second puts it faster than most, with a first answer in 23.7 s. 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
- $3.44/M
- $1.25 in · $10.00 out
- Output speed
- 121 tok/s
- measured by Artificial Analysis
- First answer
- 24 s
- Context
- 1.0M
- Full answer
- 28 s
- median, reasoning included
- Cost per run
- $0.330
- one full Intelligence Index run
- Released
- 25 Mar 2025
- Overall index54
- Reasoning51
- Coding49
- Agents & tools41
- Maths52
- Knowledge53
- Instruction following53
- Human preference63
- Multimodal64
- Long context66
- Composite47
Versions
Google has shipped 10 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 Pro | 19 Feb 2026 | 60.6 | #106 |
| Gemini 3 Pro | 18 Nov 2025 | 60.4 | #109 |
| Gemini 2.5 Prothis page | 25 Mar 2025 | 53.9 | #264 |
| Gemini 2.0 Pro | 5 Feb 2025 | 47.0 | #464 |
| Gemini 1.5 Pro 002 | 24 Sept 2024 | 41.6 | #623 |
| Gemini 1.5 Pro | 15 May 2024 | 46.1 | #495 |
| Gemini 1.5 Pro 001 | 14 May 2024 | – | – |
| Gemini 1.0 Pro 001 | 13 Dec 2023 | – | – |
| Gemini 1.0 Pro | 6 Dec 2023 | – | – |
| 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 |
|---|---|---|---|---|---|---|
| high | – | – | 121 tok/s | 24 s | $0.0035 | |
| thinking | – | – | 121 tok/s | 24 s | $0.0035 | Reasoning 36 |
| not statedbest | 53.9 | #264 | 121 tok/s | 24 s | $0.0035 | Agents & tools 41 · Coding 49 · Composite 47 · Human preference 63 · Instruction following 53 · Knowledge 53 · Long context 66 · Maths 52 · Multimodal 64 · Reasoning 51 |
How its index has moved
8 Sept 2026 to 20 Oct 2026The 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 | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| GPQA Diamondeffort not stated | – | 85.3%#91 | Epoch AI Benchmarking Hub | |
| Humanity's Last Exameffort not stated | – | 21.6%#20 | Scale AI / CAIS | |
| SimpleBencheffort not stated | – | 62.4%#21 | SimpleBench | |
| Arena Hard Promptseffort not stated | – | 1458#99 | Arena (LMArena) | |
| GPQA Diamond (AA)not in indexeffort not stated | – | 84.4%#142 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in indexeffort not stated | – | 22.5%#200 | Artificial Analysis | |
| GPQA Diamond (Vals)not in indexeffort not stated | – | 80.8%#64 | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | 70.3%#26 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | 74.9%#5 | HELM Capabilities | |
| ARC-AGI-1 | 41.0%#190 | 33.0%#205 | ARC Prize | |
| ARC-AGI-2 | 4.9%#177 | 3.8%#189 | ARC Prize | |
| CritPteffort not stated | – | 2.6%#189 | Artificial Analysis | |
| SciPredictnot in indexeffort not stated | – | 17.0%#11 | Scale AI SEAL | |
| LMCAnot in indexeffort not stated | – | 34.8%#126 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | 82.4%#120 | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | 64.4%#25 | ForecastBench |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| SWE-bench Verified (Epoch) | 57.6%#31 | Epoch AI Benchmarking Hub | |
| SciCode | 46.3%#134 | SciCode | |
| WeirdML | 54.0%#62 | WeirdML | |
| GSO-Bench | 3.9%#29 | GSO-Bench | |
| Arena Coding | 1464#123 | Arena (LMArena) | |
| Arena WebDev | 1227#131 | Arena (LMArena) | |
| SciCode (AA)not in index | 46.3%#139 | Artificial Analysis | |
| LiveCodeBench | 79.2%#71 | Vals AI | |
| SWE-bench (Vals)not in index | 54.4%#79 | Vals AI | |
| Aider Polyglot | 83.1%#3 | Aider polyglot leaderboard | |
| SWE-bench Verified (bash only) | 53.6%#32 | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | 53.6%#35 | SWE-bench | |
| Vibe Code Bench v1.1not in index | 0.4%#110 | Vals AI | |
| CadEvalnot in index | 64.0%#2 | Epoch AI Benchmarking Hub | |
| ALE-Benchnot in index | 785.5#77 | Epoch AI Benchmarking Hub | |
| AlgoTunenot in index | 1.5#11 | AlgoTune | |
| Terminal-Bench 4.0 (AA)not in index | 0.0%#152 | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | 28.5%#159 | Artificial Analysis |
Agents & tools
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| Terminal-Bencheffort not stated | – | 32.6%#58 | Terminal-Bench | |
| GDPvaleffort not stated | – | 23.3%#9 | OpenAI | |
| Remote Labor Indexeffort not stated | – | 0.8%#16 | Scale AI / CAIS | |
| Terminal-Bench Hardeffort not stated | – | 26.5%#129 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in indexeffort not stated | – | 54.1%#188 | Artificial Analysis | |
| τ³-bench Bankingnot in index | 13.7%#22 | – | τ²-bench | |
| τ³-Banking (AA)effort not stated | – | 9.7%#146 | Artificial Analysis | |
| Poker Agentnot in indexeffort not stated | – | 1032.6%#11 | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in indexeffort not stated | – | 30.3%#43 | Vals AI | |
| BALROGnot in indexeffort not stated | – | 43.3%#10 | BALROG | |
| DeepResearch Benchnot in indexeffort not stated | – | 42.8%#33 | DeepResearch Bench | |
| TheAgentCompanynot in indexeffort not stated | – | 30.3%#5 | TheAgentCompany | |
| METR Time Horizonsnot in indexeffort not stated | – | 55.4%#23 | METR | |
| Vending-Bench 2not in indexeffort not stated | – | 573.6#57 | Andon Labs | |
| AutomationBenchnot in indexeffort not stated | – | 2.2%#180 | Artificial Analysis | |
| GDP.pdf (AA)not in indexeffort not stated | – | 10.2%#132 | Artificial Analysis | |
| BrowseComp-Plusnot in indexeffort not stated | – | 19.9%#9 | BrowseComp-Plus | |
| BrowseCompnot in indexeffort not stated | – | 7.8%#4 | OpenAI | |
| DeepSearchQAnot in indexeffort not stated | – | 24.6%#9 |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FrontierMath Tiers 1–3 | 24.6%#81 | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | 0.0%#72 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 84.7%#97 | Epoch AI Benchmarking Hub | |
| MATH Level 5 | 95.9%#12 | Epoch AI Benchmarking Hub | |
| AIME (Vals) | 85.8%#35 | Vals AI | |
| IMO 2025 | 31.6%#2 | MathArena | |
| MathArena Apex | 0.5%#41 | MathArena | |
| Omni-MATH (HELM) | 41.6%#28 | HELM Capabilities | |
| MATH 500not in index | 95.2%#5 | Vals AI | |
| Arena Maths | 1439#94 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -16.4#225 | Artificial Analysis | |
| MMLU-Pro | 84.1%#65 | Vals AI | |
| LegalBench | 84.3%#35 | Vals AI | |
| CorpFin | 60.8%#63 | Vals AI | |
| TaxEval | 72.9%#55 | Vals AI | |
| MedQA | 93.1%#25 | Vals AI | |
| PRBench Finance | 38.9%#30 | Scale AI SEAL | |
| PRBench Legal | 41.4%#27 | Scale AI SEAL | |
| MultiNRC | 45.1%#15 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in index | 86.3%#4 | HELM Capabilities | |
| CaseLaw v2not in index | 63.9%#15 | Vals AI | |
| MedCodenot in index | 50.6%#17 | Vals AI | |
| MedScribenot in index | 73.5%#81 | Vals AI | |
| AA-Omniscience: accuracynot in index | 39.0%#130 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 9.1%#436 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 48.7%#182 | Artificial Analysis | |
| MultiChallenge | 53.6%#24 | Scale AI SEAL | |
| TutorBench | 55.6%#3 | Scale AI SEAL | |
| IFEval (HELM) | 84.0%#25 | HELM Capabilities | |
| WildBench (HELM) | 85.7%#6 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1438#82 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1446#85 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1443#52 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1448#89 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1456#78 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 83.8%#6 | Lech Mazur |
Multimodal
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Visioneffort not stated | – | 1263#51 | Arena (LMArena) | |
| MMMU-Proeffort not stated | – | 74.9%#107 | Artificial Analysis | |
| VISTAeffort not stated | – | 54.6%#1 | Scale AI SEAL | |
| MMMU (validation)effort not stated | – | 79.6%#7 | MMMU | |
| MMMU-Pro (official)not in indexeffort not stated | – | 68.0%#12 | MMMU | |
| MMMU-Pro (Vals)not in indexeffort not stated | – | 81.3%#40 | Vals AI | |
| MortgageTaxnot in indexeffort not stated | – | 68.9%#8 | Vals AI | |
| SAGEnot in indexeffort not stated | – | 41.9%#55 | Vals AI | |
| AudioMCnot in index | 46.9%#7 | – | Scale AI SEAL | |
| AudioMC (text output)not in index | 46.9%#6 | – | Scale AI SEAL | |
| VTBnot in indexeffort not stated | – | 11.8%#9 | Scale AI SEAL | |
| GeoBenchnot in indexeffort not stated | – | 86.0%#2 | GeoBench | |
| VPCTnot in indexeffort not stated | – | 48.0%#11 | Epoch AI Benchmarking Hub | |
| SpatialViz-Benchnot in indexeffort not stated | – | 44.7%#1 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Fiction.LiveBench 120k | 90.6%#4 | Fiction.live | |
| AA-LCR | 69.0%#232 | Artificial Analysis | |
| Arena Documentnot in index | 1431#39 | Arena (LMArena) | |
| MLCRnot in index | 12.2%#96 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | 145.1#85 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 16.1#275 | Artificial Analysis | |
| HELM Capabilities meannot in index | 74.5%#15 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 61.7%#3 | Scale AI SEAL | |
| MASKnot in index | 55.9%#44 | Scale AI SEAL | |
| PropensityBenchnot in index | 79.0%#1 | Scale AI SEAL |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
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 $0.125 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0035 | $0.0032 | 26.2 s |
| Summarize a 30-page report | 12,000 / 600 | $0.021 | $0.011 | 28.7 s |
| Code edit | 6,000 / 1,500 | $0.022 | $0.017 | 36.1 s |
| Agentic coding session | 60,000 / 4,000 | $0.115 | $0.064 | 56.8 s |
| Structured extraction | 2,000 / 200 | $0.0045 | $0.0028 | 25.3 s |
More from Google
See also
Questions people ask
- What are Gemini 2.5 Pro's benchmark scores?
- As of 24 Oct 2026, Gemini 2.5 Pro scores 53.9 ±3.2 on the BenchLeader Index, strongest in long context (66) and weakest in agents & tools (41). That rests on 53 independent benchmark results, and every score on this page links to its publisher.
- How does Gemini 2.5 Pro rank?
- Gemini 2.5 Pro is #264 of 785 ranked configurations on 24 Oct 2026. 279 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 Pro cost per 1M tokens?
- As of 24 Oct 2026, Gemini 2.5 Pro costs $1.25 per 1M input tokens and $10.00 per 1M output tokens, $3.44 blended (three input tokens per output token).
- When was Gemini 2.5 Pro released?
- Google released Gemini 2.5 Pro on 25 Mar 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 Pro: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gemini-2-5-pro.