Claude Opus 4.5 vs Gemini 3.1 Pro
Wins 0 of 8 areas
—
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Gemini 3.1 Pro is the stronger all-rounder.
Scores updated · 63 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking05Gemini 3.1 Pro5 of 5 tests
- CodingWriting and fixing software26Gemini 3.1 Pro6 of 9 tests · 1 tie
- MathCompetition and research-level math14Gemini 3.1 Pro4 of 6 tests · 1 tie
- FactsGetting facts right instead of making them up03Gemini 3.1 Pro3 of 4 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video03Gemini 3.1 Pro3 of 3 tests
- Following instructionsDoing exactly what it is asked03Gemini 3.1 Pro3 of 3 tests
- AgentsCarrying out multi-step tasks on its own13Gemini 3.1 Pro3 of 4 tests
- Long documentsFinding answers in very long texts01Gemini 3.1 Pro1 of 1 test
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Gemini 3.1 Pro costs 53% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Gemini 3.1 Pro pulls ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+69.9points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+48.9points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+39.5points ahead
Where Claude Opus 4.5 pulls ahead
- Real work tasks from 44 professionsGDPVal+32.6points ahead
- Programming problems, refreshed regularlyLiveBench · Coding+3.2points ahead
- Olympiad math problems with short answersIMOAnswerBench+3points ahead
Every test, side by side
All 63 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemini 3.1 Pro
- SWE-bench MultilingualGemini 3.1 Pro by 69.9776.9+69.9
- SciCodeGemini 3.1 Pro by 9.249.558.7+9.2
- LiveCodeBench v6Gemini 3.1 Pro by 6.984.891.7+6.9
- Terminal-Bench HardGemini 3.1 Pro by 6.84753.8+6.8
- LiveBench · Agentic CodingGemini 3.1 Pro by 4.439.744.1+4.4
- LiveBench · CodingClaude Opus 4.5 by 3.279.776.5+3.2
- LMArena · WebDevClaude Opus 4.5 by 22 rating points14681446+22 rating
- SWE-bench ProGemini 3.1 Pro by 2.25254.2+2.2
- SWE-bench Verifiedtie80.980.6tie
AgentsGemini 3.1 Pro
- BrowseCompGemini 3.1 Pro by 48.93785.9+48.9
- GDPValClaude Opus 4.5 by 32.647.314.7+32.6
- OSWorld-VerifiedGemini 3.1 Pro by 9.966.376.2+9.9
- MCP AtlasGemini 3.1 Pro by 6.962.369.2+6.9
ReasoningGemini 3.1 Pro
- ARC-AGI-2Gemini 3.1 Pro by 39.537.677.1+39.5
- Humanity's Last ExamGemini 3.1 Pro by 16.930.147+16.9
- CritPtGemini 3.1 Pro by 13.14.617.7+13.1
- GPQA DiamondGemini 3.1 Pro by 7.586.694.1+7.5
- LiveBench · ReasoningGemini 3.1 Pro by 3.980.184+3.9
FactsGemini 3.1 Pro
- SimpleQA VerifiedGemini 3.1 Pro by 27.845.773.5+27.8
- AA-Omniscience · Non-hallucinationGemini 3.1 Pro by 10.13949.1+10.1
- AA-Omniscience · AccuracyGemini 3.1 Pro by 8.346.654.9+8.3
- Vectara HHEM hallucination ratelower is bettertie10.910.4tie
Images and chartsGemini 3.1 Pro
- CharXiv (RQ)Gemini 3.1 Pro by 16.167.283.3+16.1
- MathVistaGemini 3.1 Pro by 1080.290.2+10
- MMMU-ProGemini 3.1 Pro by 8.47482.4+8.4
MathGemini 3.1 Pro
- FrontierMath Tiers 1-3 (v2)Gemini 3.1 Pro by 25.334.459.6+25.3
- FrontierMath Tier 4Gemini 3.1 Pro by 21.94.926.8+21.9
- HMMT Feb. 2026Gemini 3.1 Pro by 9.485.394.7+9.4
- AIME 2026Gemini 3.1 Pro by 3.295.198.3+3.2
- IMOAnswerBenchClaude Opus 4.5 by 38481+3
- LiveBench · Mathematicstie90.491tie
Following instructionsGemini 3.1 Pro
- IFBenchGemini 3.1 Pro by 19.15877.1+19.1
- LiveBench · Instruction FollowingGemini 3.1 Pro by 16.562.579.1+16.5
- Multi-ChallengeGemini 3.1 Pro by 12.45971.4+12.4
Other results28 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AA Agentic IndexClaude Opus 4.5 by 49.359.610.3+49.3
- Artificial Analysis Coding IndexGemini 3.1 Pro by 2147.868.8+21
- AA-OmniscienceGemini 3.1 Pro by 17.91431.9+17.9
- ARC-AGI-1Gemini 3.1 Pro by 168298+16
- MATH-VisionGemini 3.1 Pro by 12.777.189.8+12.7
- frontiermath_tier_4_v1Gemini 3.1 Pro by 12.54.216.7+12.5
- CyberGymClaude Opus 4.5 by 11.850.638.8+11.8
- NL2RepoClaude Opus 4.5 by 9.843.233.4+9.8
- MotionBenchGemini 3.1 Pro by 9.660.369.9+9.6
- Terminal-Bench 2.0Gemini 3.1 Pro by 9.259.368.5+9.2
- HLE (with tools)Gemini 3.1 Pro by 8.243.251.4+8.2
- WorldVQAGemini 3.1 Pro by 7.536.844.3+7.5
- τ²-Bench Telecom (AA run)Gemini 3.1 Pro by 6.189.595.6+6.1
- DeepSearchQA (F1)Gemini 3.1 Pro by 5.876.181.9+5.8
- ToolathlonGemini 3.1 Pro by 5.343.548.8+5.3
- LiveCodeBenchGemini 3.1 Pro by 4.78791.7+4.7
- Finance AgentGemini 3.1 Pro by 4.555.259.7+4.5
- livebench_data_analysisGemini 3.1 Pro by 4.174.478.5+4.1
- livebench_languageGemini 3.1 Pro by 4.181.385.4+4.1
- LiveBenchGemini 3.1 Pro by 3.97679.9+3.9
- MMMLUGemini 3.1 Pro by 1.890.892.6+1.8
- HMMT Nov. 2025Gemini 3.1 Pro by 1.593.394.8+1.5
- MMLU-ProGemini 3.1 Pro by 1.589.591+1.5
- vectara_answer_ratetie98.799.4tie
- vectara_avg_summary_lengthtie114.5107.7tie
- AA Intelligencetie29.129.7tie
- vectara_factual_consistencytie89.189.6tie
- SimpleVQAtie69.769.9tie
Questions people ask
Which is better, Claude Opus 4.5 or Gemini 3.1 Pro?
Gemini 3.1 Pro wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Claude Opus 4.5 wins none.
Which is better for coding?
Gemini 3.1 Pro. It wins 6 of the 9 coding tests both models report; Claude Opus 4.5 wins 2, and 1 is a tie.
Which is cheaper?
Claude Opus 4.5 costs $5.00 per million input tokens and $25.00 per million output tokens; Gemini 3.1 Pro costs $2.00 and $12.00. That makes Gemini 3.1 Pro about 53% cheaper for the same work.
How do you compare the two?
We use the 63 benchmark tests both models have published scores on. The verdict counts the 35 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 28 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.