Claude Opus 4.7 vs GPT-5.6 Luna
Wins 5 of 8 areas
Coding · Reasoning · Facts · Images and charts · Following instructions
Wins 2 of 8 areas
Math · Long documents
Claude Opus 4.7 is the stronger all-rounder.GPT-5.6 Luna is cheaper and better at math.
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.
- Images and chartsUnderstanding pictures, charts and video40Claude Opus 4.74 of 5 tests · 1 tie
- CodingWriting and fixing software41Claude Opus 4.74 of 7 tests · 2 ties
- ReasoningHard problems that need careful thinking41Claude Opus 4.74 of 7 tests · 2 ties
- FactsGetting facts right instead of making them up30Claude Opus 4.73 of 3 tests
- Following instructionsDoing exactly what it is asked10Claude Opus 4.71 of 1 test
- MathCompetition and research-level math14GPT-5.6 Luna4 of 5 tests
- Long documentsFinding answers in very long texts02GPT-5.6 Luna2 of 2 tests
- AgentsCarrying out multi-step tasks on its own33Even3 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.
GPT-5.6 Luna costs 95% 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+50.3points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+16.2points ahead
- Common-sense trick questionsSimpleBench+14.9points ahead
Where GPT-5.6 Luna pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+26.8points ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+15.5points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+11.9points ahead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
- SWE-bench VerifiedGPT-5.6 Luna by 5.487.693+5.4
- LMArena · WebDevClaude Opus 4.7 by 38 rating points15571519+38 rating
- LiveBench · Agentic CodingClaude Opus 4.7 by 2.350.748.4+2.3
- Terminal-Bench 2.1Claude Opus 4.7 by 2.283.180.9+2.2
- SWE-bench ProClaude Opus 4.7 by 1.664.362.7+1.6
- SciCodetie54.553.6tie
- LiveBench · Codingtie82.182.9tie
AgentsEven
- AA IT-Bench SREClaude Opus 4.7 by 6.446.740.3+6.4
- GDPValGPT-5.6 Luna by 5.442.648+5.4
- OSWorld-VerifiedClaude Opus 4.7 by 5.47872.6+5.4
- BrowseCompGPT-5.6 Luna by 479.383.3+4
- τ-Bench V3 · BankingClaude Opus 4.7 by 3.534.631.1+3.5
- AA ApexAgentsGPT-5.6 Luna by 1.933.935.8+1.9
ReasoningClaude Opus 4.7
- ARC-AGI-2Claude Opus 4.7 by 16.275.859.6+16.2
- SimpleBenchClaude Opus 4.7 by 14.961.746.8+14.9
- CritPtGPT-5.6 Luna by 8.61220.6+8.6
- Humanity's Last ExamClaude Opus 4.7 by 2.842.339.5+2.8
- LiveBench · ReasoningClaude Opus 4.7 by 1.687.285.6+1.6
- GPQA Diamondtie91.491.1tie
- ARC-AGI-3tie0.20.2tie
FactsClaude Opus 4.7
- AA-Omniscience · Non-hallucinationClaude Opus 4.7 by 50.357.77.4+50.3
- SimpleQA VerifiedClaude Opus 4.7 by 10.751.741+10.7
- AA-Omniscience · AccuracyClaude Opus 4.7 by 6.248.942.7+6.2
Images and chartsClaude Opus 4.7
- CharXiv (RQ)Claude Opus 4.7 by 8.39182.7+8.3
- MathVistaClaude Opus 4.7 by 7.584.476.9+7.5
- LMArena · VisionClaude Opus 4.7 by 58 rating points13141256+58 rating
- BLINKClaude Opus 4.7 by 1.870.468.6+1.8
- MMMU-Protie78.878.6tie
MathGPT-5.6 Luna
- FrontierMath Tier 4GPT-5.6 Luna by 26.831.758.5+26.8
- FrontierMath Tiers 1-3 (v2)GPT-5.6 Luna by 11.970.282.1+11.9
- LiveBench · MathematicsClaude Opus 4.7 by 5.792.887.2+5.7
- HMMT Feb. 2026GPT-5.6 Luna by 4.693.998.5+4.6
- AIME 2026GPT-5.6 Luna by 1.895.897.6+1.8
Long documentsGPT-5.6 Luna
- MRCR v2 8 needle 128k (average)GPT-5.6 Luna by 15.559.374.8+15.5
- AA-LCRGPT-5.6 Luna by 578.783.7+5
Following instructionsClaude Opus 4.7
- LiveBench · Instruction FollowingClaude Opus 4.7 by 6.666.760.1+6.6
Other results14 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.
- AA-OmniscienceClaude Opus 4.7 by 37.627.3-10.3+37.6
- DeepSWEGPT-5.6 Luna by 13.25467.2+13.2
- ToolathlonClaude Opus 4.7 by 5.959.353.4+5.9
- HLE (with tools)Claude Opus 4.7 by 5.854.748.9+5.8
- livebench_languageClaude Opus 4.7 by 5.377.972.6+5.3
- ERQAGPT-5.6 Luna by 452.556.5+4
- AA IntelligenceClaude Opus 4.7 by 3.440.737.3+3.4
- SimpleVQAClaude Opus 4.7 by 3.356.553.2+3.3
- AA Agentic IndexGPT-5.6 Luna by 3.239.542.7+3.2
- RealWorldQAClaude Opus 4.7 by 3.275.672.4+3.2
- Artificial Analysis Coding IndexClaude Opus 4.7 by 2.273.671.4+2.2
- EmbSpatial-BenchClaude Opus 4.7 by 1.777.275.5+1.7
- ARC-AGI-1Claude Opus 4.7 by 1.39290.7+1.3
- livebench_data_analysistie78.378tie
Questions people ask
Which is better, Claude Opus 4.7 or GPT-5.6 Luna?
Claude Opus 4.7 wins five of the eight areas we test: coding, reasoning, facts, images and charts and following instructions. GPT-5.6 Luna wins math and long documents, and costs 95% less. They are level on agents.
Which is better for coding?
Claude Opus 4.7. It wins 4 of the 7 coding tests both models report; GPT-5.6 Luna wins 1, and 2 are ties.
Which is cheaper?
Claude Opus 4.7 costs $5.00 per million input tokens and $25.00 per million output tokens; GPT-5.6 Luna costs $0.20 and $1.20. That makes GPT-5.6 Luna about 95% 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 36 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 14 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.