Models / Gemini 2.5 Pro
Gemini 2.5 Pro
Gemini 2.5 Pro is strongest at writing; most buyers will do better with one of the alternatives below. The cheapest model of similar intelligence is Qwen3.5-9B, at $0.11 per million tokens against $3.44.
Superseded. Gemini 3.1 Pro Preview is Google's newer model in this line, at $4.50 blended: see Gemini 3.1 Pro Preview or compare the two.
Retiring 1 Dec 2026. Google retires gemini-2.5-pro; the official replacement is gemini-3.8-flash. How to re-test before you switch.
Results 16 Nov 2026 · 29 results · 6 sources
- Price per million tokens
- $1.25 in · $10.00 out
Google's list price, 10 Nov 2026 · compare prices$0.12 cached input - Speed
- Not measured
- Intelligence Index
- 16.1
Artificial Analysis - Spring Prompt overall
- Not ranked yet
needs our benchmarks and two groups of results - Developer
- Google, based in the US
Weights not published - Released
- 25 Mar 2025
1,048,576 tokens of context
Where it stands among the models you could choose
Available models grouped by price, intelligence, speed and developer location. Retired models are excluded.
| Among | Intelligence Index | Spring Prompt overall | Price, blended | Speed |
|---|---|---|---|---|
| Models available today | 109th of 185Claude Opus 5.5 | – | $3.44 blended, 79th percentilecheapest: Mistral Nemo | – |
| Models priced $3–10 per million tokens | 31st of 38Claude Opus 5.5 | – | sets this group | – |
| Models with similar intelligence | sets this group | – | $3.44 blended, 96th percentilecheapest: Qwen3.5-9B | – |
| Models from developers based in the US | 60th of 92Claude Opus 5.5 | – | $3.44 blended, 61st percentilecheapest: Gemma 3 4B | – |
The name under each rank is the leader of that group; green is the top quarter of the group and red the bottom quarter, by rank. Price is shown as a percentile: the share of the group that costs less, so lower is cheaper. "Sets this group" marks the measure the group is defined by. Spring Prompt overall is our score out of 100 across every source (how it works). Price is the developer's list price (or the typical OpenRouter host where we haven't read one) for three input tokens to one output; intelligence and speed are from Artificial Analysis (data sourced from Artificial Analysis); "similar intelligence" means within 4 points on its Intelligence Index. Developer locations are where each company is based, not where a model is served. Groups under 3 models are not ranked.
How good is it, and for what?
Each use case ranks the available models its benchmarks measured, then those priced $3–10 per million tokens. The bar shows its percentile in that field, best to the right. Open a row for the results behind it.
Product listingsTurning a sparse product feed and photos into listings that can go live
- CatalogBench: has not measured this model.
Decks from an analysisTurning a finished analysis into a deck you could present as it is
- DeckBench: has not measured this model.
User surveysPlanning a user survey and reading its results without being misled
- SurveyBench: has not measured this model.
Marketing planningPlanning a year of ad spend without overspending
- ROASBench: has not measured this model.
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls
| Benchmark | Gemini 2.5 Pro | Rank among available | Best available |
|---|---|---|---|
| Artificial AnalysisArtificial Analysis: Terminal-Bench 4.0 · reported | 0.0% | 67th of 107 | Kimi K3.1 Pro 67.2% |
| Artificial AnalysisArtificial Analysis: Terminal-Bench 2.1 · reported | 28.5% | 78th of 107 | Claude Fable 5.1 91.4% |
| Artificial AnalysisArtificial Analysis: τ-bench banking · reported | 9.7% | 73rd of 104 | Grok 4.6 50.7% |
| Vending-Bench 2Vending-Bench 2: money after a year · reported | $573.64 | 43rd of 53 | GPT-6 Astra $15,515 |
- tau2-bench: has not measured this model.
- Berkeley Function Calling Leaderboard (BFCL) V4: has not measured this model.
- OpenHands Index: has not measured this model.
- Microsoft STATE-Bench: has not measured this model.
Professional workReal tasks from banking, consulting and law, business documents and freelance projects
- APEX-Agents: has not measured this model.
- GDP.pdf: has not measured this model.
- Remote Labor Index: has not measured this model.
Reasoning and knowledgeHard questions across science, maths and general knowledge
| Benchmark | Gemini 2.5 Pro | Rank among available | Best available |
|---|---|---|---|
| Artificial AnalysisArtificial Analysis: Artificial Analysis Intelligence Index · reported | 16.1 | 109th of 185 | Claude Opus 5.5 57.6 |
| Artificial AnalysisArtificial Analysis: Humanity's Last Exam · reported | 22.5% | 96th of 186 | Claude Opus 5.5 61.4% |
| Artificial AnalysisArtificial Analysis: GPQA Diamond · reported | 84.4% | 75th of 168 | GPT-6 Astra 96.3% |
WritingWhat people prefer in blind comparisons, and judged writing quality
| Benchmark | Gemini 2.5 Pro | Rank among available | Best available |
|---|---|---|---|
| Arena (formerly LMArena)Arena Text: overall · reported | 1,446 | 56th of 153 | Gemini 4 Argon 1,525 |
| UGI LeaderboardUGI: writing score · reported | 65.2 | 23rd of 124 | Gemini 3.8 Flash 78.6 |
Sticking to the factsSummarising without inventing things, and factual answers
| Benchmark | Gemini 2.5 Pro | Rank among available | Best available |
|---|---|---|---|
| Vectara Hallucination LeaderboardVectara: hallucination rate · reported | 7.0% | 22nd of 71 | GPT-5.4 nano 3.1% |
| Arena (formerly LMArena)Arena Text factuality: overall · reported | 1,462 | 34th of 109 | Gemini 4 Argon 1,511 |
- SimpleQA Verified (Epoch AI): has not measured this model.
Speed under pressureGood decisions against a real clock (fast chess)
- BulletBench: has not measured this model.
Nearest alternatives
- Cheaper at similar intelligenceQwen3.5-9B$0.11 blended · Intelligence Index 13.3Compare with Gemini 2.5 Pro →
- Google's newer model in this lineGemini 3.1 Pro Preview$4.50 blended · Intelligence Index 29.7Compare with Gemini 2.5 Pro →
- Open weights at similar intelligenceQwen3.5-9B$0.11 blended · Intelligence Index 13.3Compare with Gemini 2.5 Pro →
Against Gemini 2.5 Pro at $3.44 blended, Intelligence Index 16.1. "Similar intelligence" means within 4 points or better.
Where to run it
1 host on OpenRouter at the standard service tier, with a median of $3.44 blended per million tokens. Cheapest first.
| Host | Quantisation | Input | Output | Blended |
|---|---|---|---|---|
| Google AI Studio | not stated | $1.25 | $10.00 | $3.44 |
Other service tiers, not counted above: Google AI Studio flex $0.62 in, $5.00 out; Google AI Studio priority $2.25 in, $18.00 out. Flex and batch are cheaper but slower or queued; priority and fast cost more.
Prices per million tokens from OpenRouter's public endpoint list, checked 10 Nov 2026.
Against the alternatives
The models you would most likely weigh it against: the leaders of the groups above.
Scroll sideways to see every alternative.
| Gemini 2.5 Pro | Claude Opus 5.5 most intelligent at $3–10 per million tokens | Qwen3.5-9B cheapest with similar intelligence | Claude Fable 5.2 leads overall | Gemini 4 Argon Google's best other model | |
|---|---|---|---|---|---|
| Price per million tokens | $3.44 | $8.00 | $0.11 | $16.00 | $4.00 |
| Tokens a second | – | 96 | 110 | – | – |
| Intelligence Index | 16.1 | 57.6 | 13.3 | – | 52.6 |
| Spring Prompt overall | – | 82 | – | 90 | 72 |
| Rank among available models, by use case | |||||
| Product listings | – | 13th | – | 5th | 11th |
| Decks from an analysis | – | 7th | – | 4th | 23rd |
| User surveys | – | 4th | – | 1st | 15th |
| Marketing planning | – | 9th | – | 1st | 13th |
| Agents and tool use | 118th of 159 | 3rd | 111th | – | 5th |
| Professional work | – | 3rd | – | – | – |
| Reasoning and knowledge | 102nd of 187 | 1st | 116th | – | 4th |
| Writing | 44th of 169 | 5th | 112th | – | 1st |
| Sticking to the facts | 39th of 139 | 3rd | – | – | 1st |
| Speed under pressure | – | – | – | 8th | 26th |
| Side by side → | Side by side → | Side by side → | |||
Shaded figures are better than Gemini 2.5 Pro's. A dash means no result.
Research and examples
Where it has been measured
31st most intelligent of the 38 models priced $3–10 per million tokens. Among models available today, in the bottom quarter for agents and tool use.
Artificial Analysis lists it but has not published a speed for it yet
- Arena (formerly LMArena)9 results
- Artificial Analysis12 results
- tau2-bench2 results
- UGI Leaderboard3 results
- Vectara Hallucination Leaderboard2 results
- Vending-Bench 21 result
- BulletBenchNot measured
- CatalogBenchNot measured
- DeckBenchNot measured
- ROASBenchNot measured
- SurveyBenchNot measured
- APEX-AgentsNot measured
- Berkeley Function Calling Leaderboard (BFCL) V4Not measured
- GDP.pdfNot measured
- Microsoft STATE-BenchNot measured
- OpenHands IndexNot measured
- Remote Labor IndexNot measured
- SimpleQA Verified (Epoch AI)Not measured
Every result
29 published results from 6 sources, each in the source's own units, with the configuration that produced it.
Arena (formerly LMArena) · reported by Arena (formerly LMArena) · 9 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| Arena Document: overallArena rating, higher is better | 1,431 1,424–1,437 |
39 of 44 | Gemini 2.5 Pro | 13 Sep 2026 |
| Arena Search: overallArena rating, higher is better | 1,142 1,137–1,147 |
29 of 34 | Gemini 2.5 Pro (arena search) | 24 Aug 2026 |
| Arena Text factuality: overallArena rating, higher is better | 1,462 1,459–1,465 |
45 of 182 | Gemini 2.5 Pro | 2 Oct 2026 |
| Arena Text: business, management and financeArena rating, higher is better | 1,437 1,433–1,442 |
106 of 406 | Gemini 2.5 Pro | 2 Oct 2026 |
| Arena Text: creative writingArena rating, higher is better | 1,443 1,438–1,448 |
54 of 411 | Gemini 2.5 Pro | 2 Oct 2026 |
| Arena Text: expert promptsArena rating, higher is better | 1,458 1,452–1,465 |
104 of 363 | Gemini 2.5 Pro | 2 Oct 2026 |
| Arena Text: instruction followingArena rating, higher is better | 1,438 1,434–1,442 |
82 of 413 | Gemini 2.5 Pro | 2 Oct 2026 |
| Arena Text: overallArena rating, higher is better | 1,446 1,443–1,448 |
86 of 413 | Gemini 2.5 Pro | 2 Oct 2026 |
| Arena Text: writing, literature and languageArena rating, higher is better | 1,445 1,440–1,449 |
63 of 412 | Gemini 2.5 Pro | 2 Oct 2026 |
Not shown: Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
Artificial Analysis · reported by Artificial Analysis · 12 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| Artificial Analysis: Artificial Analysis Coding Indexindex score, higher is better | 33.3 | 131 of 187 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better | 16.1 | 244 of 413 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: GPQA Diamond% of questions, higher is better | 84.4% | 131 of 360 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: Humanity's Last Exam% of questions, higher is better | 22.5% | 188 of 419 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: IFBench% of instructions, higher is better | 48.7% | 146 of 288 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better | 69.0% | 210 of 407 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: SciCode% of problems, higher is better | 46.3% | 140 of 205 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: Terminal-Bench 2.1% of tasks, higher is better | 28.5% | 137 of 188 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: Terminal-Bench 4.0% of tasks, higher is better | 0.0% | 152 of 204 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better | 26.5% | 121 of 282 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: τ-bench banking% of tasks, higher is better | 9.7% | 126 of 175 | Gemini 2.5 Pro | 20 Oct 2026 |
| Artificial Analysis: τ²-bench telecom% of tasks, higher is better | 54.1% | 156 of 286 | Gemini 2.5 Pro | 20 Oct 2026 |
Not shown: Not your codebase or tools.
tau2-bench · reported by tau2-bench · 2 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| tau2 v1.0.1, banking knowledge, alltools retrieval: consistency (pass^4)% of tasks, higher is better | 1.0% | 21 of 21 | Gemini 2.5 Pro (high reasoning, tau2) | 5 May 2026 |
| tau2 v1.0.1, banking knowledge, alltools retrieval: task success (pass^1)% of tasks, higher is better | 13.7% | 20 of 21 | Gemini 2.5 Pro (high reasoning, tau2) | 5 May 2026 |
Not shown: Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.
UGI Leaderboard · reported by UGI Leaderboard · 3 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| UGI: requested-length error% off the requested word count, lower is better | 19.0% | 171 of 379 | Gemini 2.5 Pro | 9 Sep 2025 |
| UGI: style adherencescore from 0 to 1, higher is better | 0.40 | 29 of 379 | Gemini 2.5 Pro | 9 Sep 2025 |
| UGI: writing scorescore out of 100, higher is better | 65.2 | 64 of 379 | Gemini 2.5 Pro | 9 Sep 2025 |
Not shown: Not other format limits such as character counts or bullet counts.
Vectara Hallucination Leaderboard · reported by Vectara Hallucination Leaderboard · 2 results
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| Vectara: answer rate% of documents, higher is better | 99.1% | 64 of 108 | Gemini 2.5 Pro | 22 Sep 2026 |
| Vectara: hallucination rate% of summaries, lower is better | 7.0% | 32 of 108 | Gemini 2.5 Pro | 22 Sep 2026 |
Not shown: Not a quality score: a low rate usually means content filters were triggered, and hallucination rates are measured on answered documents only.
Vending-Bench 2 · reported by Vending-Bench 2 · 1 result
| Measure | Value | Rank | Configuration | Dated |
|---|---|---|---|---|
| Vending-Bench 2: money after a yearUS dollars, higher is better | $573.64 | 51 of 63 | Gemini 2.5 Pro | 1 Oct 2026 |
Not shown: Not a real business; one simulated market with set rules.
Test your own prompt.
Benchmarks aren't your task. Run your prompt on the models you could use, with every answer checked and costed.