Claude Opus 4.5 vs Gemini 3 Pro
Wins 2 of 8 areas
Coding · Agents
Wins 3 of 8 areas
Reasoning · Images and charts · Following instructions
Gemini 3 Pro wins more areas, narrowly.Claude Opus 4.5 is better at coding.
Scores updated · 83 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.
- Images and chartsUnderstanding pictures, charts and video04Gemini 3 Pro4 of 4 tests
- ReasoningHard problems that need careful thinking13Gemini 3 Pro3 of 4 tests
- Following instructionsDoing exactly what it is asked02Gemini 3 Pro2 of 2 tests
- CodingWriting and fixing software43Claude Opus 4.54 of 7 tests
- AgentsCarrying out multi-step tasks on its own21Claude Opus 4.52 of 3 tests
- FactsGetting facts right instead of making them up22Even2 each
- MathCompetition and research-level math11Even1 each · 1 tie
- Long documentsFinding answers in very long texts11Even1 each
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 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 Pro pulls ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+58points ahead
- Short factual questions, answered correctlySimpleQA Verified+27.2points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+22.2points ahead
Where Claude Opus 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+30.5points ahead
- Real work tasks from 44 professionsGDPVal+13.1points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+8.7points ahead
Every test, side by side
All 83 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.5
- SWE-bench MultilingualGemini 3 Pro by 58765+58
- SWE-bench ProClaude Opus 4.5 by 8.75243.3+8.7
- SciCodeGemini 3 Pro by 6.649.556.1+6.6
- Terminal-Bench HardClaude Opus 4.5 by 5.34741.7+5.3
- SWE-bench VerifiedClaude Opus 4.5 by 4.780.976.2+4.7
- LMArena · WebDevClaude Opus 4.5 by 29 rating points14681439+29 rating
- LiveCodeBench v6Gemini 3 Pro by 2.684.887.4+2.6
AgentsClaude Opus 4.5
- BrowseCompGemini 3 Pro by 22.23759.2+22.2
- GDPValClaude Opus 4.5 by 13.147.334.2+13.1
- MCP AtlasClaude Opus 4.5 by 8.262.354.1+8.2
ReasoningGemini 3 Pro
- Humanity's Last ExamGemini 3 Pro by 9.630.139.7+9.6
- ARC-AGI-2Claude Opus 4.5 by 6.537.631.1+6.5
- CritPtGemini 3 Pro by 4.54.69.1+4.5
- GPQA DiamondGemini 3 Pro by 4.286.690.8+4.2
FactsEven
- AA-Omniscience · Non-hallucinationClaude Opus 4.5 by 30.5398.5+30.5
- SimpleQA VerifiedGemini 3 Pro by 27.245.772.9+27.2
- AA-Omniscience · AccuracyGemini 3 Pro by 9.246.655.8+9.2
- Vectara HHEM hallucination ratelower is betterClaude Opus 4.5 by 2.710.913.6+2.7
Images and chartsGemini 3 Pro
- CharXiv (RQ)Gemini 3 Pro by 14.267.281.4+14.2
- MathVistaGemini 3 Pro by 9.680.289.8+9.6
- MMMU-ProGemini 3 Pro by 6.27480.2+6.2
- OCRBenchGemini 3 Pro by 3.886.590.3+3.8
MathEven
- AIME 2026Claude Opus 4.5 by 3.495.191.7+3.4
- HMMT Feb. 2026Gemini 3 Pro by 1.185.386.4+1.1
- IMOAnswerBenchtie8483.1tie
Long documentsEven
- LongBench v2Gemini 3 Pro by 3.864.468.2+3.8
- AA-LCRClaude Opus 4.5 by 1.377.376+1.3
Following instructionsGemini 3 Pro
- IFBenchGemini 3 Pro by 12.45870.4+12.4
- Multi-ChallengeGemini 3 Pro by 6.75965.7+6.7
Other results54 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.
- InfoVQA (val)Claude Opus 4.5 by 19.776.957.2+19.7
- VIBE-BackendClaude Opus 4.5 by 19.39878.7+19.3
- WideSearch (item-f1)Claude Opus 4.5 by 19.276.257+19.2
- FinSearchCompT2&T3Claude Opus 4.5 by 16.366.249.9+16.3
- frontiermath_tier_4_v1Gemini 3 Pro by 14.64.218.8+14.6
- VIBE-iOSClaude Opus 4.5 by 14.29075.8+14.2
- OJ-Bench (cpp)Gemini 3 Pro by 13.954.668.5+13.9
- VIBE-AndroidClaude Opus 4.5 by 13.592.278.7+13.5
- OctoCodingbenchClaude Opus 4.5 by 13.336.222.9+13.3
- DeepSearchQA (F1)Claude Opus 4.5 by 12.976.163.2+12.9
- WorldVQAGemini 3 Pro by 10.636.847.4+10.6
- MotionBenchGemini 3 Pro by 1060.370.3+10
- MATH-VisionGemini 3 Pro by 977.186.1+9
- VIBE (Average)Claude Opus 4.5 by 8.390.782.4+8.3
- AA Agentic IndexClaude Opus 4.5 by 7.659.652+7.6
- Multi-SWE-BenchClaude Opus 4.5 by 7.35042.7+7.3
- AIME 2025Gemini 3 Pro by 7.292.8100+7.2
- ToolathlonClaude Opus 4.5 by 7.143.536.4+7.1
- ARC-AGI-1Claude Opus 4.5 by 78275+7
- ARC-AGI-2 (Verified)Claude Opus 4.5 by 6.537.631.1+6.5
- VIBE-SimulationGemini 3 Pro by 5.28489.2+5.2
- Terminal-Bench 2.0Claude Opus 4.5 by 5.159.354.2+5.1
- AIME25 no toolsGemini 3 Pro by 59196+5
- LiveCodeBenchGemini 3 Pro by 58792+5
- metr_re_benchClaude Opus 4.5 by 54035+5
- ZeroBenchGemini 3 Pro by 538+5
- HMMT Feb. 2025Gemini 3 Pro by 4.692.997.5+4.6
- τ²-Bench (Retail)Claude Opus 4.5 by 3.688.985.3+3.6
- VideoMMMUGemini 3 Pro by 3.284.487.6+3.2
- ZeroBench (w/ tools)Gemini 3 Pro by 3912+3
- AIME 2026 I (Tools-allowed)Claude Opus 4.5 by 2.793.390.6+2.7
- vectara_factual_consistencyClaude Opus 4.5 by 2.789.186.4+2.7
- HLE (with tools)Gemini 3 Pro by 2.643.245.8+2.6
- LiveBenchClaude Opus 4.5 by 2.67673.4+2.6
- τ²-Bench Telecom (AA run)Claude Opus 4.5 by 2.489.587.1+2.4
- Seal-0Claude Opus 4.5 by 2.247.745.5+2.2
- metr_hcastClaude Opus 4.5 by 2.173.471.3+2.1
- SWE-PerfGemini 3 Pro by 1.84.76.5+1.8
- BrowseComp (context management)Gemini 3 Pro by 1.457.859.2+1.4
- AA-OmniscienceGemini 3 Pro by 1.31415.3+1.3
- Artificial Analysis Coding IndexClaude Opus 4.5 by 1.347.846.5+1.3
- vectara_avg_summary_lengthClaude Opus 4.5 by 12.6 rating points114.5101.9+12.6 rating
- AA IntelligenceClaude Opus 4.5 by 1.129.128+1.1
- MMMLUGemini 3 Pro by 190.891.8+1
- OmniDocBench 1.5tie87.788.5tie
- vectara_answer_ratetie98.799.4tie
- MMLU-Protie89.590.1tie
- SWT-benchtie80.279.7tie
- VIBE-Webtie89.189.5tie
- MMVUtie77.377.5tie
- τ²-Benchtie98.298tie
- HMMT Nov. 2025tie93.393.3tie
- metr_swaatie99.599.5tie
- SimpleVQAtie69.769.7tie
Questions people ask
Which is better, Claude Opus 4.5 or Gemini 3 Pro?
Gemini 3 Pro wins three of the eight areas we test: reasoning, images and charts and following instructions. Claude Opus 4.5 wins coding and agents. They are level on facts, math and long documents.
Which is better for coding?
Claude Opus 4.5. It wins 4 of the 7 coding tests both models report; Gemini 3 Pro wins 3.
Which is cheaper?
Claude Opus 4.5 costs $5.00 per million input tokens and $25.00 per million output tokens; Gemini 3 Pro costs $2.00 and $12.00. That makes Gemini 3 Pro about 53% cheaper for the same work.
How do you compare the two?
We use the 83 benchmark tests both models have published scores on. The verdict counts the 29 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 54 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.