Claude Sonnet 4.5 vs Gemini 3 Pro
Wins 0 of 8 areas
—
Wins 7 of 8 areas
Coding · Agents · Reasoning · Images and charts · Math · Long documents · Following instructions
Gemini 3 Pro is the stronger all-rounder.
Scores updated · 74 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 Pro5 of 5 tests
- AgentsCarrying out multi-step tasks on its own03Gemini 3 Pro3 of 3 tests
- Images and chartsUnderstanding pictures, charts and video03Gemini 3 Pro3 of 3 tests
- CodingWriting and fixing software24Gemini 3 Pro4 of 7 tests · 1 tie
- Long documentsFinding answers in very long texts02Gemini 3 Pro2 of 2 tests
- Following instructionsDoing exactly what it is asked02Gemini 3 Pro2 of 2 tests
- MathCompetition and research-level math01Gemini 3 Pro1 of 1 test
- FactsGetting facts right instead of making them up22Even2 each
Math 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 Pro costs 22% 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
- Short factual questions, answered correctlySimpleQA Verified+42.2points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+35.1points ahead
- Recent programming contest problemsLiveCodeBench v6+23.4points ahead
Where Claude Sonnet 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+42.4points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+2points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+1points ahead
Every test, side by side
All 74 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemini 3 Pro
- LiveCodeBench v6Gemini 3 Pro by 23.46487.4+23.4
- SciCodeGemini 3 Pro by 10.445.756.1+10.4
- Terminal-Bench HardGemini 3 Pro by 6.135.641.7+6.1
- LMArena · WebDevGemini 3 Pro by 46 rating points13941440+46 rating
- SWE-bench MultilingualClaude Sonnet 4.5 by 26765+2
- SWE-bench VerifiedClaude Sonnet 4.5 by 177.276.2+1
- SWE-bench Protie43.643.3tie
AgentsGemini 3 Pro
- BrowseCompGemini 3 Pro by 35.124.159.2+35.1
- GDPValGemini 3 Pro by 13.720.534.2+13.7
- MCP AtlasGemini 3 Pro by 10.343.854.1+10.3
ReasoningGemini 3 Pro
- SimpleBenchGemini 3 Pro by 22.154.376.4+22.1
- Humanity's Last ExamGemini 3 Pro by 21.917.839.7+21.9
- ARC-AGI-2Gemini 3 Pro by 17.513.631.1+17.5
- CritPtGemini 3 Pro by 81.19.1+8
- GPQA DiamondGemini 3 Pro by 7.483.490.8+7.4
FactsEven
- AA-Omniscience · Non-hallucinationClaude Sonnet 4.5 by 42.450.98.5+42.4
- SimpleQA VerifiedGemini 3 Pro by 42.230.772.9+42.2
- AA-Omniscience · AccuracyGemini 3 Pro by 22.932.955.8+22.9
- Vectara HHEM hallucination ratelower is betterClaude Sonnet 4.5 by 1.61213.6+1.6
Images and chartsGemini 3 Pro
- CharXiv (RQ)Gemini 3 Pro by 14.267.281.4+14.2
- MMMU-ProGemini 3 Pro by 11.568.780.2+11.5
- MathVistaGemini 3 Pro by 1079.889.8+10
Long documentsGemini 3 Pro
- LongBench v2Gemini 3 Pro by 6.461.868.2+6.4
- AA-LCRGemini 3 Pro by 3.772.376+3.7
Following instructionsGemini 3 Pro
- IFBenchGemini 3 Pro by 13.157.370.4+13.1
- Multi-ChallengeGemini 3 Pro by 10.455.365.7+10.4
Other results47 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.
- OJ-Bench (cpp)Gemini 3 Pro by 38.130.468.5+38.1
- AA Agentic IndexGemini 3 Pro by 34.517.552+34.5
- MRCRGemini 3 Pro by 34.355.489.7+34.3
- BrowseComp (context management)Gemini 3 Pro by 33.126.159.2+33.1
- HMMT Feb. 2025Gemini 3 Pro by 3067.597.5+30
- LiveCodeBenchGemini 3 Pro by 217192+21
- HarmBenchClaude Sonnet 4.5 by 19.59272.5+19.5
- ARC-AGI-2 (Verified)Gemini 3 Pro by 17.513.631.1+17.5
- Arena-Hard (Creative Writing)Gemini 3 Pro by 16.976.793.6+16.9
- AIR-Bench 2024Claude Sonnet 4.5 by 16.689.873.2+16.6
- AA-OmniscienceGemini 3 Pro by 15.4-0.115.3+15.4
- frontiermath_tier_4_v1Gemini 3 Pro by 14.64.218.8+14.6
- HLE (with tools)Gemini 3 Pro by 13.83245.8+13.8
- AIME 2025Gemini 3 Pro by 1387100+13
- SimpleVQAGemini 3 Pro by 12.157.669.7+12.1
- VIBE-BackendClaude Sonnet 4.5 by 12.190.878.7+12.1
- HMMT Nov. 2025Gemini 3 Pro by 11.681.793.3+11.6
- ARC-AGI-1Gemini 3 Pro by 11.363.775+11.3
- SWT-benchGemini 3 Pro by 10.269.579.7+10.2
- VIBE-SimulationGemini 3 Pro by 10.179.189.2+10.1
- Arena-Hard (Hard Prompt)Gemini 3 Pro by 9.363.372.6+9.3
- τ²-Bench Telecom (AA run)Gemini 3 Pro by 978.187.1+9
- VIBE-AndroidClaude Sonnet 4.5 by 8.887.578.7+8.8
- matharena_visual_math_overallGemini 3 Pro by 8.475.884.2+8.4
- AIME25 no toolsGemini 3 Pro by 88896+8
- AA IntelligenceGemini 3 Pro by 7.320.728+7.3
- Artificial Analysis Coding IndexClaude Sonnet 4.5 by 5.652.146.5+5.6
- VIBE-iOSClaude Sonnet 4.5 by 5.481.275.8+5.4
- ToolathlonClaude Sonnet 4.5 by 4.64136.4+4.6
- Terminal-Bench 2.0Gemini 3 Pro by 4.25054.2+4.2
- vectara_answer_rateGemini 3 Pro by 3.895.699.4+3.8
- SWE-PerfGemini 3 Pro by 3.536.5+3.5
- VIBE (Average)Claude Sonnet 4.5 by 2.885.282.4+2.8
- MMMLUGemini 3 Pro by 2.789.191.8+2.7
- OmniDocBench 1.5Gemini 3 Pro by 2.785.888.5+2.7
- MMLU-ProGemini 3 Pro by 2.687.590.1+2.6
- vectara_avg_summary_lengthClaude Sonnet 4.5 by 25.9 rating points127.8101.9+25.9 rating
- simple_safety_testsClaude Sonnet 4.5 by 2.510097.5+2.5
- VIBE-WebGemini 3 Pro by 2.287.389.5+2.2
- Multi-SWE-BenchClaude Sonnet 4.5 by 1.644.342.7+1.6
- vectara_factual_consistencyClaude Sonnet 4.5 by 1.68886.4+1.6
- anthropic_red_teamClaude Sonnet 4.5 by 1.198.297+1.1
- XSTesttie96.497.3tie
- τ²-Bench (Retail)tie86.285.3tie
- bbqtie98.998.4tie
- OctoCodingbenchtie22.822.9tie
- τ²-Benchtie9898tie
Questions people ask
Which is better, Claude Sonnet 4.5 or Gemini 3 Pro?
Gemini 3 Pro wins seven of the eight areas we test: coding, agents, reasoning, images and charts, math, long documents and following instructions. Claude Sonnet 4.5 wins none. They are level on facts.
Which is better for coding?
Gemini 3 Pro. It wins 4 of the 7 coding tests both models report; Claude Sonnet 4.5 wins 2, and 1 is a tie.
Which is cheaper?
Claude Sonnet 4.5 costs $3.00 per million input tokens and $15.00 per million output tokens; Gemini 3 Pro costs $2.00 and $12.00. That makes Gemini 3 Pro about 22% cheaper for the same work.
How do you compare the two?
We use the 74 benchmark tests both models have published scores on. The verdict counts the 27 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 47 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.