Compare / Gemini 2.5 Pro vs Gemini 3.1 Pro Preview
Gemini 2.5 Pro vs Gemini 3.1 Pro Preview
Which is cheaper, which is more capable, and which does better at each kind of work, from 29 results on 6 sources that measured both models.
Of the 29 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.
Results as of 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee).
Gemini 2.5 Pro or Gemini 3.1 Pro Preview? The short answer
Pick Gemini 3.1 Pro Preview if quality matters most: it scores higher (Intelligence Index 29.7 against 16.1). Pick Gemini 2.5 Pro if cost matters more: it costs 24% less per token.
- Gemini 2.5 Pro is cheaper: $3.44 against $4.50 per million tokens, 24% less.
- Gemini 3.1 Pro Preview scores higher on the Artificial Analysis Intelligence Index: 29.7 against 16.1.
- Gemini 3.1 Pro Preview is newer: released 19 Feb 2026, against 25 Mar 2025 for Gemini 2.5 Pro.
- On third-party benchmarks, on the results both models have: Gemini 3.1 Pro Preview does better for agents and tool use, reasoning and knowledge and writing.
Head to head
Every fact side by side
| Gemini 2.5 Pro | Gemini 3.1 Pro Preview | |
|---|---|---|
| Price per million tokens | $1.25 in · $10.00 out | $2.00 in · $12.00 out |
| Blended (3 in : 1 out) | $3.44 | $4.50 |
| Cheapest host, blended | – | $4.50 (Google) |
| Cached input, per million | $0.12 cached input | $0.20 cached input |
| Intelligence Index | 16.1 | 29.7 |
| Spring Prompt overall | – | 46 |
| Output speed, tokens a second | – | 130 |
| Context window | 1,048,576 tokens | 1,048,576 tokens |
| Released | 25 Mar 2025 | 19 Feb 2026 (newer) |
| Weights | Closed | Closed |
| Developer based in | the US | the US |
Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.
Which is better for what
Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.
From our research
- We Told 11 AI Models What Happened in 2026. Most of Them Thought We Were Making It Up. 12 Oct 2026
- Google Gemini 3 Review: The Benchmarks Actually Match the Hype 20 Nov 2025
- GPT-5.1 First Look: Smarter, Warmer… But Not a Breakthrough 13 Nov 2025
- Gemini 4 Argon review: Opus-level product data at a quarter of the cost, until you ask it to sell 14 Oct 2026
Every shared result
Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.
| Measure | Gemini 2.5 Pro | Gemini 3.1 Pro Preview |
|---|---|---|
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results | ||
| Overall · TextArena rating, higher is better 74 · 15 of 387 measured | 1,446 | 1,487 |
| Overall · Text factualityArena rating, higher is better 38 · 21 of 163 measured | 1,462 | 1,472 |
| Business, management and finance · TextArena rating, higher is better 92 · 23 of 380 measured | 1,437 | 1,477 |
| Artificial Analysis reported by Artificial Analysis · 3 results | ||
| Artificial Analysis Intelligence Indexindex score, higher is better 128 · 57 of 249 measured | 16.1 | 29.7 |
| GPQA Diamond% of questions, higher is better 87 · 5 of 232 measured | 84.4% | 94.1% |
| Humanity's Last Exam% of questions, higher is better 108 · 19 of 250 measured | 22.5% | 47.0% |
| τ²-bench reported by τ²-bench · 2 results | ||
| Consistency (pass^4) · tau2 v1.0.1, banking knowledge, alltools retrieval% of tasks, higher is better 21 · 17 of 21 measured | 1.0%high reasoning, tau2 | 9.3%high reasoning, tau2 |
| Task success (pass^1) · tau2 v1.0.1, banking knowledge, alltools retrieval% of tasks, higher is better 20 · 15 of 21 measured | 13.7%high reasoning, tau2 | 26.0%high reasoning, tau2 |
| UGI Leaderboard reported by UGI Leaderboard · 3 results | ||
| Writing score · UGIscore out of 100, higher is better 28 · 5 of 230 measured | 65.2 | 72.2low reasoning |
| Requested-length error · UGI% off the requested word count, lower is better 99 · 99 of 230 measured | 19.0% | 19.0%low reasoning |
| Style adherence · UGIscore from 0 to 1, higher is better 16 · 122 of 230 measured | 0.40 | 0.34low reasoning |
| Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results | ||
| Hallucination rate · Vectara% of summaries, lower is better 32 · 61 of 104 measured | 7.0% | 10.4% |
| Answer rate · Vectara% of documents, higher is better 61 · 55 of 104 measured | 99.1% | 99.4% |
| Vending-Bench 2 reported by Vending-Bench 2 · 1 result | ||
| Money after a yearUS dollars, higher is better 46 · 45 of 58 measured | $573.64 | $911.21 |
Show all 29 results (15 more)
| Measure | Gemini 2.5 Pro | Gemini 3.1 Pro Preview |
|---|---|---|
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 6 results | ||
| Creative writing · TextArena rating, higher is better 45 · 10 of 385 measured | 1,443 | 1,480 |
| Expert prompts · TextArena rating, higher is better 91 · 26 of 338 measured | 1,458 | 1,509 |
| Instruction following · TextArena rating, higher is better 70 · 17 of 387 measured | 1,438 | 1,480 |
| Overall · DocumentArena rating, higher is better 34 · 23 of 38 measured | 1,431 | 1,459 |
| Overall · SearchArena rating, higher is better 29 · 6 of 33 measured | 1,142arena search | 1,208arena search |
| Writing, literature and language · TextArena rating, higher is better 53 · 11 of 386 measured | 1,445 | 1,480 |
| Artificial Analysis reported by Artificial Analysis · 9 results | ||
| Terminal-Bench 2.1% of tasks, higher is better 85 · 36 of 131 measured | 28.5% | 73.8% |
| Terminal-Bench 4.0% of tasks, higher is better 72 · 51 of 120 measured | 0.0% | 4.0% |
| τ-bench banking% of tasks, higher is better 79 · 46 of 121 measured | 9.7% | 21.4% |
| Artificial Analysis Coding Indexindex score, higher is better 81 · 30 of 130 measured | 33.3 | 68.8 |
| IFBench% of instructions, higher is better 108 · 13 of 195 measured | 48.7% | 77.1% |
| Long-context reasoning (AA-LCR)% of questions, higher is better 123 · 27 of 241 measured | 69.0% | 82.0% |
| SciCode% of problems, higher is better 69 · 14 of 121 measured | 46.3% | 58.7% |
| Terminal-Bench Hard% of tasks, higher is better 74 · 8 of 190 measured | 26.5% | 53.8% |
| τ²-bench telecom% of tasks, higher is better 99 · 16 of 193 measured | 54.1% | 95.6% |
Compare them with other models
Quick answers
Is Gemini 2.5 Pro better than Gemini 3.1 Pro Preview?
Pick Gemini 3.1 Pro Preview if quality matters most: it scores higher (Intelligence Index 29.7 against 16.1). Pick Gemini 2.5 Pro if cost matters more: it costs 24% less per token.
Which is cheaper, Gemini 2.5 Pro or Gemini 3.1 Pro Preview?
Gemini 2.5 Pro is cheaper: $3.44 against $4.50 per million tokens, 24% less.
Which is better for agents and tool use, Gemini 2.5 Pro or Gemini 3.1 Pro Preview?
Gemini 3.1 Pro Preview ranks 66 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 118 for Gemini 2.5 Pro.
Which is better for reasoning and knowledge, Gemini 2.5 Pro or Gemini 3.1 Pro Preview?
Gemini 3.1 Pro Preview ranks 29 of 187 models on sale for hard questions across science, maths and general knowledge, against 102 for Gemini 2.5 Pro.
Which is better for writing, Gemini 2.5 Pro or Gemini 3.1 Pro Preview?
Gemini 3.1 Pro Preview ranks 9 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 44 for Gemini 2.5 Pro.
Which is better for sticking to the facts, Gemini 2.5 Pro or Gemini 3.1 Pro Preview?
Gemini 3.1 Pro Preview ranks 33 of 139 models on sale for summarising without inventing things, and factual answers, against 39 for Gemini 2.5 Pro.
Run Gemini 2.5 Pro and Gemini 3.1 Pro Preview on your own prompt.
Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.