Claude Opus 4.6 vs Claude Opus 4.7
Wins 1 of 8 areas
Long documents
Wins 6 of 8 areas
Coding · Reasoning · Facts · Images and charts · Math · Following instructions
Claude Opus 4.7 is the stronger all-rounder.Claude Opus 4.6 is better at long documents.
Scores updated · 73 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 software08Claude Opus 4.78 of 8 tests
- MathCompetition and research-level math14Claude Opus 4.74 of 6 tests · 1 tie
- FactsGetting facts right instead of making them up03Claude Opus 4.73 of 4 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video03Claude Opus 4.73 of 4 tests · 1 tie
- Following instructionsDoing exactly what it is asked02Claude Opus 4.72 of 2 tests
- ReasoningHard problems that need careful thinking23Claude Opus 4.73 of 7 tests · 2 ties
- Long documentsFinding answers in very long texts10Claude Opus 4.61 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own22Even2 each · 1 tie
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.
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+20.5points ahead
- Multi-step tasks using many real software toolsMCP Atlas+14.6points ahead
- Reasoning about charts from research papersCharXiv (RQ)+13.6points ahead
Where Claude Opus 4.6 pulls ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+24.7points ahead
- Real work tasks from 44 professionsGDPVal+13.1points ahead
- Common-sense trick questionsSimpleBench+5.9points ahead
Every test, side by side
All 73 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.7
- SWE-bench ProClaude Opus 4.7 by 10.953.464.3+10.9
- SWE-bench VerifiedClaude Opus 4.7 by 6.880.887.6+6.8
- Terminal-Bench HardClaude Opus 4.7 by 5.346.251.5+5.3
- LiveBench · CodingClaude Opus 4.7 by 3.978.282.1+3.9
- SWE-bench MultilingualClaude Opus 4.7 by 2.777.880.5+2.7
- SciCodeClaude Opus 4.7 by 2.651.954.5+2.6
- LiveBench · Agentic CodingClaude Opus 4.7 by 1.74950.7+1.7
- LMArena · WebDevClaude Opus 4.7 by 12 rating points15461558+12 rating
AgentsEven
- MCP AtlasClaude Opus 4.7 by 14.662.777.3+14.6
- GDPValClaude Opus 4.6 by 13.155.942.8+13.1
- OSWorld-VerifiedClaude Opus 4.7 by 5.372.778+5.3
- BrowseCompClaude Opus 4.6 by 4.78479.3+4.7
- AA ApexAgentstie3333.9tie
ReasoningClaude Opus 4.7
- ARC-AGI-2Claude Opus 4.7 by 768.875.8+7
- SimpleBenchClaude Opus 4.6 by 5.967.661.7+5.9
- Humanity's Last ExamClaude Opus 4.7 by 2.439.942.3+2.4
- GPQA DiamondClaude Opus 4.7 by 1.889.691.4+1.8
- LiveBench · ReasoningClaude Opus 4.6 by 1.588.787.2+1.5
- CritPttie12.612tie
- ARC-AGI-3tie0.50.2tie
FactsClaude Opus 4.7
- AA-Omniscience · Non-hallucinationClaude Opus 4.7 by 20.537.257.7+20.5
- SimpleQA VerifiedClaude Opus 4.7 by 4.74751.7+4.7
- AA-Omniscience · AccuracyClaude Opus 4.7 by 1.94748.9+1.9
- Vectara HHEM hallucination ratelower is bettertie12.212tie
Images and chartsClaude Opus 4.7
- CharXiv (RQ)Claude Opus 4.7 by 13.677.491+13.6
- OCRBenchv2Claude Opus 4.7 by 8.548.456.9+8.5
- MMMU-ProClaude Opus 4.7 by 3.475.478.8+3.4
- LMArena · Visiontie13161313tie
MathClaude Opus 4.7
- FrontierMath Tier 4Claude Opus 4.7 by 4.926.831.7+4.9
- FrontierMath Tiers 1-3 (v2)Claude Opus 4.7 by 4.26670.2+4.2
- LiveBench · MathematicsClaude Opus 4.7 by 3.689.392.8+3.6
- USAMO 2026Claude Opus 4.7 by 3.166.269.3+3.1
- HMMT Feb. 2026Claude Opus 4.6 by 2.396.293.9+2.3
- AIME 2026tie96.795.8tie
Long documentsClaude Opus 4.6
- MRCR v2 8 needle 128k (average)Claude Opus 4.6 by 24.78459.3+24.7
- AA-LCRtie7878.7tie
Following instructionsClaude Opus 4.7
- IFBenchClaude Opus 4.7 by 5.553.158.6+5.5
- LiveBench · Instruction FollowingClaude Opus 4.7 by 3.463.366.7+3.4
Other results35 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.
- XBOW visual-acuity benchmarkClaude Opus 4.7 by 4454.598.5+44
- AA Agentic IndexClaude Opus 4.6 by 28.167.639.5+28.1
- Artificial Analysis Coding IndexClaude Opus 4.7 by 25.548.173.6+25.5
- OfficeQA ProClaude Opus 4.7 by 23.557.180.6+23.5
- ScreenSpot-Pro (No tools)Claude Opus 4.7 by 21.857.779.5+21.8
- AA-OmniscienceClaude Opus 4.7 by 13.613.727.3+13.6
- GDPval-AA (Elo)Claude Opus 4.7 by 134 rating points16191753+134 rating
- OfficeQAClaude Opus 4.7 by 12.873.586.3+12.8
- ERQAClaude Opus 4.7 by 11.740.852.5+11.7
- Toolathlon Avg turnsClaude Opus 4.7 by 916.925.9+9
- AA IntelligenceClaude Opus 4.7 by 8.831.940.7+8.8
- livebench_data_analysisClaude Opus 4.7 by 8.469.978.3+8.4
- BabyVisionClaude Opus 4.7 by 7.414.822.2+7.4
- SWE-bench MultimodalClaude Opus 4.7 by 7.427.134.5+7.4
- CharXiv Reasoning (With tools)Claude Opus 4.7 by 6.384.791+6.3
- Toolathlon Pass@∞Claude Opus 4.7 by 5.647.252.8+5.6
- livebench_languageClaude Opus 4.6 by 5.483.377.9+5.4
- ScreenSpot-Pro (With tools)Claude Opus 4.7 by 4.583.187.6+4.5
- Terminal-Bench 2.0Claude Opus 4.7 by 465.469.4+4
- Finance AgentClaude Opus 4.7 by 3.760.764.4+3.7
- τ²-Bench Telecom (AA run)Claude Opus 4.6 by 3.592.188.6+3.5
- HiL-Bench (Tools-allowed)Claude Opus 4.7 by 3.438.341.7+3.4
- Legal Agent BenchmarkClaude Opus 4.7 by 2.94.27.1+2.9
- ToolathlonClaude Opus 4.7 by 2.556.859.3+2.5
- vectara_answer_rateClaude Opus 4.6 by 1.899.898+1.8
- RealWorldQAClaude Opus 4.7 by 1.773.975.6+1.7
- HLE (with tools)Claude Opus 4.7 by 1.453.354.7+1.4
- vectara_avg_summary_lengthClaude Opus 4.7 by 11.5 rating points137.6149.1+11.5 rating
- ARC-AGI-1Claude Opus 4.6 by 19392+1
- BigLaw Benchtie90.290.9tie
- CyberGymtie73.873.1tie
- LiveBenchtie76.376.9tie
- MMMLUtie91.191.5tie
- vectara_factual_consistencytie87.888tie
- frontiermath_tier_4_v1tie22.922.9tie
Questions people ask
Which is better, Claude Opus 4.6 or Claude Opus 4.7?
Claude Opus 4.7 wins six of the eight areas we test: coding, reasoning, facts, images and charts, math and following instructions. Claude Opus 4.6 wins long documents. They are level on agents.
Which is better for coding?
Claude Opus 4.7. It wins 8 of the 8 coding tests both models report; Claude Opus 4.6 wins none.
Which is cheaper?
Claude Opus 4.6 costs $5.00 per million input tokens and $25.00 per million output tokens; Claude Opus 4.7 costs $5.00 and $25.00. For a million tokens read plus a million written, they cost about the same.
How do you compare the two?
We use the 73 benchmark tests both models have published scores on. The verdict counts the 38 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 35 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.