GPT-5.1 vs GPT-5.6 Luna
Wins 1 of 6 areas
Facts
Wins 5 of 6 areas
Coding · Agents · Reasoning · Images and charts · Long documents
GPT-5.6 Luna is the stronger all-rounder.GPT-5.1 is better at facts.
Scores updated · 29 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. 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 software04GPT-5.6 Luna4 of 4 tests
- ReasoningHard problems that need careful thinking14GPT-5.6 Luna4 of 5 tests
- AgentsCarrying out multi-step tasks on its own03GPT-5.6 Luna3 of 3 tests
- Images and chartsUnderstanding pictures, charts and video01GPT-5.6 Luna1 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts01GPT-5.6 Luna1 of 1 test
- FactsGetting facts right instead of making them up21GPT-5.12 of 3 tests
Long documents 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 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 GPT-5.6 Luna pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+42points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+32.5points ahead
- Real work tasks from 44 professionsGDPVal+31.7points ahead
Where GPT-5.1 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+40.7points ahead
- Short factual questions, answered correctlySimpleQA Verified+7points ahead
- Common-sense trick questionsSimpleBench+6.4points ahead
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.6 Luna
- Terminal-Bench 2.1GPT-5.6 Luna by 28.552.480.9+28.5
- SWE-bench VerifiedGPT-5.6 Luna by 16.776.393+16.7
- LMArena · WebDevGPT-5.6 Luna by 123 rating points13961519+123 rating
- SciCodeGPT-5.6 Luna by 10.343.353.6+10.3
AgentsGPT-5.6 Luna
- BrowseCompGPT-5.6 Luna by 32.550.883.3+32.5
- GDPValGPT-5.6 Luna by 31.716.348+31.7
- τ-Bench V3 · BankingGPT-5.6 Luna by 15.215.931.1+15.2
ReasoningGPT-5.6 Luna
- ARC-AGI-2GPT-5.6 Luna by 4217.659.6+42
- CritPtGPT-5.6 Luna by 15.74.920.6+15.7
- Humanity's Last ExamGPT-5.6 Luna by 1128.539.5+11
- SimpleBenchGPT-5.1 by 6.453.246.8+6.4
- GPQA DiamondGPT-5.6 Luna by 3.887.391.1+3.8
FactsGPT-5.1
- AA-Omniscience · Non-hallucinationGPT-5.1 by 40.748.17.4+40.7
- SimpleQA VerifiedGPT-5.1 by 74841+7
- AA-Omniscience · AccuracyGPT-5.6 Luna by 537.742.7+5
Images and chartsGPT-5.6 Luna
- MMMU-ProGPT-5.6 Luna by 3.175.578.6+3.1
- LMArena · Visiontie12491256tie
Other results11 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.
- FrontierMath (overall)GPT-5.6 Luna by 51.926.778.6+51.9
- Artificial Analysis Coding IndexGPT-5.6 Luna by 2249.471.4+22
- AA Agentic IndexGPT-5.6 Luna by 21.121.642.7+21.1
- ARC-AGI-1GPT-5.6 Luna by 17.972.890.7+17.9
- HealthBench Professional length-adjustedGPT-5.6 Luna by 16.139.655.7+16.1
- AA-OmniscienceGPT-5.1 by 15.75.4-10.3+15.7
- AA IntelligenceGPT-5.6 Luna by 12.624.737.3+12.6
- HealthBench Hard length-adjustedGPT-5.6 Luna by 6.625.432+6.6
- HLE (with tools)GPT-5.6 Luna by 6.242.748.9+6.2
- HealthBench length-adjustedGPT-5.6 Luna by 4.950.955.8+4.9
- HealthBench Consensus length-adjustedtie9595.1tie
Questions people ask
Which is better, GPT-5.1 or GPT-5.6 Luna?
GPT-5.6 Luna wins five of the six areas where both have results: coding, agents, reasoning, images and charts and long documents. GPT-5.1 wins facts.
Which is better for coding?
GPT-5.6 Luna. It wins 4 of the 4 coding tests both models report; GPT-5.1 wins none.
Which is cheaper?
GPT-5.1 costs $1.25 per million input tokens and $10.00 per million output tokens; GPT-5.6 Luna costs $0.20 and $1.20. That makes GPT-5.6 Luna about 88% cheaper for the same work.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 18 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 11 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.