Claude Opus 4.7 vs Gemini 3 Flash Preview
Wins 5 of 8 areas
Coding · Agents · Reasoning · Images and charts · Math
Wins 2 of 8 areas
Long documents · Following instructions
Claude Opus 4.7 is the stronger all-rounder.Gemini 3 Flash Preview is cheaper and better at long documents.
Scores updated · 50 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 software60Claude Opus 4.76 of 6 tests
- AgentsCarrying out multi-step tasks on its own50Claude Opus 4.75 of 5 tests
- ReasoningHard problems that need careful thinking40Claude Opus 4.74 of 5 tests · 1 tie
- MathCompetition and research-level math30Claude Opus 4.73 of 4 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video21Claude Opus 4.72 of 3 tests
- Long documentsFinding answers in very long texts01Gemini 3 Flash Preview1 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked01Gemini 3 Flash Preview1 of 1 test
- FactsGetting facts right instead of making them up22Even2 each
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.
Gemini 3 Flash Preview costs 88% 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 Opus 4.7 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+50.7points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+42.2points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+25.1points ahead
Where Gemini 3 Flash Preview pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+19.4points ahead
- Short factual questions, answered correctlySimpleQA Verified+15.1points ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+7.9points ahead
Every test, side by side
All 50 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.7
- Terminal-Bench 2.1Claude Opus 4.7 by 25.183.158+25.1
- SWE-bench ProClaude Opus 4.7 by 14.764.349.6+14.7
- Terminal-Bench HardClaude Opus 4.7 by 12.951.538.6+12.9
- LMArena · WebDevClaude Opus 4.7 by 118 rating points15571439+118 rating
- SWE-bench VerifiedClaude Opus 4.7 by 9.687.678+9.6
- SciCodeClaude Opus 4.7 by 3.954.550.6+3.9
AgentsClaude Opus 4.7
- MCP AtlasClaude Opus 4.7 by 19.977.357.4+19.9
- τ-Bench V3 · BankingClaude Opus 4.7 by 13.834.620.8+13.8
- OSWorld-VerifiedClaude Opus 4.7 by 12.97865.1+12.9
- GDPValClaude Opus 4.7 by 7.642.835.2+7.6
- AA ApexAgentsClaude Opus 4.7 by 6.233.927.7+6.2
ReasoningClaude Opus 4.7
- ARC-AGI-2Claude Opus 4.7 by 42.275.833.6+42.2
- Humanity's Last ExamClaude Opus 4.7 by 5.742.336.6+5.7
- CritPtClaude Opus 4.7 by 3.4128.6+3.4
- GPQA DiamondClaude Opus 4.7 by 1.691.489.8+1.6
- SimpleBenchtie61.761.1tie
FactsEven
- AA-Omniscience · Non-hallucinationClaude Opus 4.7 by 50.757.77+50.7
- SimpleQA VerifiedGemini 3 Flash Preview by 15.151.766.8+15.1
- AA-Omniscience · AccuracyGemini 3 Flash Preview by 4.548.953.4+4.5
- Vectara HHEM hallucination ratelower is betterClaude Opus 4.7 by 1.51213.5+1.5
Images and chartsClaude Opus 4.7
- CharXiv (RQ)Claude Opus 4.7 by 10.79180.3+10.7
- LMArena · VisionClaude Opus 4.7 by 28 rating points13131285+28 rating
- MMMU-ProGemini 3 Flash Preview by 1.178.879.9+1.1
MathClaude Opus 4.7
- FrontierMath Tiers 1-3 (v2)Claude Opus 4.7 by 1970.251.2+19
- FrontierMath Tier 4Claude Opus 4.7 by 14.631.717.1+14.6
- HMMT Feb. 2026Claude Opus 4.7 by 4.593.989.4+4.5
- AIME 2026tie95.896.7tie
Long documentsGemini 3 Flash Preview
- MRCR v2 8 needle 128k (average)Gemini 3 Flash Preview by 7.959.367.2+7.9
- AA-LCRtie78.778tie
Following instructionsGemini 3 Flash Preview
- IFBenchGemini 3 Flash Preview by 19.458.678+19.4
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 (Elo)Claude Opus 4.7 by 549 rating points17531204+549 rating
- Artificial Analysis Coding IndexClaude Opus 4.7 by 3173.642.6+31
- Blueprint-Bench 2Claude Opus 4.7 by 24.524.50+24.5
- Terminal-Bench 2.0Claude Opus 4.7 by 21.869.447.6+21.8
- frontiermath_tier_4_v1Claude Opus 4.7 by 18.722.94.2+18.7
- AA-OmniscienceClaude Opus 4.7 by 17.227.310.1+17.2
- AA IntelligenceClaude Opus 4.7 by 14.440.726.3+14.4
- Humanity’s Last ExamClaude Opus 4.7 by 13.246.933.7+13.2
- ScreenSpot-Pro (No tools)Claude Opus 4.7 by 10.479.569.1+10.4
- AA Agentic IndexGemini 3 Flash Preview by 10.239.549.7+10.2
- ToolathlonClaude Opus 4.7 by 9.959.349.4+9.9
- Finance Agent v2Claude Opus 4.7 by 8.951.542.5+8.9
- τ²-Bench Telecom (AA run)Claude Opus 4.7 by 8.288.680.4+8.2
- ARC-AGI-1Claude Opus 4.7 by 7.39284.7+7.3
- Legal Agent BenchmarkClaude Opus 4.7 by 7.17.10+7.1
- vectara_avg_summary_lengthClaude Opus 4.7 by 58.9 rating points149.190.2+58.9 rating
- LiveBenchClaude Opus 4.7 by 4.576.972.4+4.5
- vectara_answer_rateGemini 3 Flash Preview by 1.89899.8+1.8
- vectara_factual_consistencyClaude Opus 4.7 by 1.58886.5+1.5
- MMMLUtie91.591.8tie
Questions people ask
Which is better, Claude Opus 4.7 or Gemini 3 Flash Preview?
Claude Opus 4.7 wins five of the eight areas we test: coding, agents, reasoning, images and charts and math. Gemini 3 Flash Preview wins long documents and following instructions, and costs 88% less. They are level on facts.
Which is better for coding?
Claude Opus 4.7. It wins 6 of the 6 coding tests both models report; Gemini 3 Flash Preview wins none.
Which is cheaper?
Claude Opus 4.7 costs $5.00 per million input tokens and $25.00 per million output tokens; Gemini 3 Flash Preview costs $0.50 and $3.00. That makes Gemini 3 Flash Preview about 88% cheaper for the same work.
How do you compare the two?
We use the 50 benchmark tests both models have published scores on. The verdict counts the 30 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.