Claude Sonnet 5 vs Gemini 3.1 Pro
Wins 4 of 8 areas
Coding · Agents · Math · Long documents
Wins 4 of 8 areas
Reasoning · Facts · Images and charts · Following instructions
The two are evenly matched.Claude Sonnet 5 is better at coding and math; Gemini 3.1 Pro at reasoning and facts.
Scores updated · 51 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.
- CodingWriting and fixing software71Claude Sonnet 57 of 8 tests
- MathCompetition and research-level math40Claude Sonnet 54 of 4 tests
- AgentsCarrying out multi-step tasks on its own41Claude Sonnet 54 of 5 tests
- Long documentsFinding answers in very long texts10Claude Sonnet 51 of 2 tests · 1 tie
- ReasoningHard problems that need careful thinking13Gemini 3.1 Pro3 of 5 tests · 1 tie
- FactsGetting facts right instead of making them up12Gemini 3.1 Pro2 of 3 tests
- Images and chartsUnderstanding pictures, charts and video12Gemini 3.1 Pro2 of 3 tests
- Following instructionsDoing exactly what it is asked01Gemini 3.1 Pro1 of 1 test
Following instructions 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.
Claude Sonnet 5 costs 14% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Claude Sonnet 5 pulls ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+55.2points ahead
- Real work tasks from 44 professionsGDPVal+33.6points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+15.9points ahead
Where Gemini 3.1 Pro pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+39.8points ahead
- Common-sense trick questionsSimpleBench+19points ahead
- Follows detailed instructions exactlyLiveBench · Instruction Following+15.2points ahead
Every test, side by side
All 51 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 5
- LiveBench · Agentic CodingClaude Sonnet 5 by 15.359.444.1+15.3
- LMArena · WebDevClaude Sonnet 5 by 93 rating points15391446+93 rating
- SWE-bench ProClaude Sonnet 5 by 963.254.2+9
- Terminal-Bench 2.1Claude Sonnet 5 by 6.780.573.8+6.7
- SWE-bench VerifiedClaude Sonnet 5 by 4.685.280.6+4.6
- SciCodeGemini 3.1 Pro by 4.454.358.7+4.4
- LiveBench · CodingClaude Sonnet 5 by 4.280.776.5+4.2
- SWE-bench MultilingualClaude Sonnet 5 by 1.478.376.9+1.4
AgentsClaude Sonnet 5
- GDPValClaude Sonnet 5 by 33.648.314.7+33.6
- τ-Bench V3 · BankingClaude Sonnet 5 by 15.937.321.4+15.9
- Terminal-Bench 4.0Claude Sonnet 5 by 10.114.14+10.1
- OSWorld-VerifiedClaude Sonnet 5 by 581.276.2+5
- BrowseCompGemini 3.1 Pro by 1.284.785.9+1.2
ReasoningGemini 3.1 Pro
- SimpleBenchGemini 3.1 Pro by 1960.679.6+19
- Humanity's Last ExamGemini 3.1 Pro by 5.741.347+5.7
- LiveBench · ReasoningClaude Sonnet 5 by 4.788.784+4.7
- GPQA DiamondGemini 3.1 Pro by 391.194.1+3
- CritPttie16.917.7tie
FactsGemini 3.1 Pro
- SimpleQA VerifiedGemini 3.1 Pro by 39.833.773.5+39.8
- AA-Omniscience · AccuracyGemini 3.1 Pro by 14.84054.9+14.8
- AA-Omniscience · Non-hallucinationClaude Sonnet 5 by 11.560.649.1+11.5
Images and chartsGemini 3.1 Pro
- MMMU-ProGemini 3.1 Pro by 5.177.382.4+5.1
- CharXiv (RQ)Claude Sonnet 5 by 588.383.3+5
- LMArena · VisionGemini 3.1 Pro by 21 rating points12751296+21 rating
MathClaude Sonnet 5
- FrontierMath Tiers 1-3 (v2)Claude Sonnet 5 by 5.965.659.6+5.9
- USAMO 2026Claude Sonnet 5 by 5.179.574.4+5.1
- FrontierMath Tier 4Claude Sonnet 5 by 2.529.326.8+2.5
- LiveBench · MathematicsClaude Sonnet 5 by 1.992.991+1.9
Long documentsClaude Sonnet 5
- MRCR v2 8 needle 128k (average)Claude Sonnet 5 by 55.281.526.3+55.2
- AA-LCRtie8282tie
Following instructionsGemini 3.1 Pro
- LiveBench · Instruction FollowingGemini 3.1 Pro by 15.263.979.1+15.2
Other results20 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.
- GDPval-AA v2Claude Sonnet 5 by 656 rating points1618962+656 rating
- GDP (Surge AI)Claude Sonnet 5 by 64.981.616.7+64.9
- GDPval-AA v2 EloClaude Sonnet 5 by 619 rating points1584965+619 rating
- DeepSWE 1.1Claude Sonnet 5 by 425412+42
- Program BenchClaude Sonnet 5 by 37.877.339.5+37.8
- AA Agentic IndexClaude Sonnet 5 by 3444.310.3+34
- MLE-BenchClaude Sonnet 5 by 24.366.942.6+24.3
- AA-OmniscienceGemini 3.1 Pro by 15.416.431.9+15.4
- OfficeQA ProGemini 3.1 Pro by 13.159.472.5+13.1
- Terminal-Bench 2.0Claude Sonnet 5 by 11.980.468.5+11.9
- livebench_languageGemini 3.1 Pro by 10.47585.4+10.4
- AA IntelligenceClaude Sonnet 5 by 8.538.229.7+8.5
- livebench_data_analysisGemini 3.1 Pro by 6.871.778.5+6.8
- HLE (with tools)Claude Sonnet 5 by 657.451.4+6
- Legal Agent BenchmarkClaude Sonnet 5 by 5.85.80+5.8
- ToolathlonClaude Sonnet 5 by 5.554.348.8+5.5
- CharXiv Reasoning (With tools)Claude Sonnet 5 by 5.188.383.2+5.1
- Artificial Analysis Coding IndexClaude Sonnet 5 by 2.771.568.8+2.7
- LVBenchClaude Sonnet 5 by 2.368.566.2+2.3
- ProteinGym Hardtie36.637tie
Questions people ask
Which is better, Claude Sonnet 5 or Gemini 3.1 Pro?
Claude Sonnet 5 and Gemini 3.1 Pro each win four of the eight areas we test. Claude Sonnet 5 wins coding, agents, math and long documents; Gemini 3.1 Pro wins reasoning, facts, images and charts and following instructions. Claude Sonnet 5 costs 14% less.
Which is better for coding?
Claude Sonnet 5. It wins 7 of the 8 coding tests both models report; Gemini 3.1 Pro wins 1.
Which is cheaper?
Claude Sonnet 5 costs $2.00 per million input tokens and $10.00 per million output tokens; Gemini 3.1 Pro costs $2.00 and $12.00. That makes Claude Sonnet 5 about 14% cheaper for the same work.
How do you compare the two?
We use the 51 benchmark tests both models have published scores on. The verdict counts the 31 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 20 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.