GPT-5.6 Luna vs GPT-6 Luna
Wins 3 of 8 areas
Reasoning · Math · Following instructions
Wins 2 of 8 areas
Coding · Facts
GPT-5.6 Luna wins more areas, narrowly.GPT-6 Luna is cheaper and better at coding.
Scores updated · 37 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.
- ReasoningHard problems that need careful thinking30GPT-5.6 Luna3 of 5 tests · 2 ties
- MathCompetition and research-level math21GPT-5.6 Luna2 of 3 tests
- Following instructionsDoing exactly what it is asked10GPT-5.6 Luna1 of 1 test
- CodingWriting and fixing software13GPT-6 Luna3 of 4 tests
- FactsGetting facts right instead of making them up02GPT-6 Luna2 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own11Even1 each
- Images and chartsUnderstanding pictures, charts and video11Even1 each
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
Long documents and following instructions rest on a single test each.
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-6 Luna costs 57% 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
- Follows detailed instructions exactlyLiveBench · Instruction Following+4.2points ahead
- Programming problems, refreshed regularlyLiveBench · Coding+3.9points ahead
- Fresh logic puzzles that change regularlyLiveBench · Reasoning+3.8points ahead
Where GPT-6 Luna pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+15.9points ahead
- Fixes real GitHub issues, working as an agentLiveBench · Agentic Coding+2.8points ahead
- Fresh competition math problemsLiveBench · Mathematics+1.9points ahead
Every test, side by side
All 37 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-6 Luna
- LMArena · WebDevGPT-6 Luna by 58 rating points15191577+58 rating
- LiveBench · CodingGPT-5.6 Luna by 3.982.979+3.9
- LiveBench · Agentic CodingGPT-6 Luna by 2.848.451.2+2.8
- SciCodeGPT-6 Luna by 153.654.6+1
AgentsEven
- GDPValGPT-5.6 Luna by 1.44846.6+1.4
- Terminal-Bench 4.0GPT-6 Luna by 111.612.6+1
ReasoningGPT-5.6 Luna
- LiveBench · ReasoningGPT-5.6 Luna by 3.885.681.8+3.8
- CritPtGPT-5.6 Luna by 1.220.619.4+1.2
- Humanity's Last ExamGPT-5.6 Luna by 139.538.5+1
- ARC-AGI-2tie59.659.3tie
- ARC-AGI-3tie0.20.1tie
FactsGPT-6 Luna
- AA-Omniscience · Non-hallucinationGPT-6 Luna by 15.97.423.3+15.9
- AA-Omniscience · AccuracyGPT-6 Luna by 1.142.743.8+1.1
- SimpleQA Verifiedtie4141.4tie
Images and chartsEven
- LMArena · VisionGPT-5.6 Luna by 45 rating points12561211+45 rating
- MMMU-ProGPT-6 Luna by 1.178.679.7+1.1
MathGPT-5.6 Luna
- FrontierMath Tiers 1-3 (v2)GPT-5.6 Luna by 3.182.179+3.1
- FrontierMath Tier 4GPT-5.6 Luna by 2.458.556.1+2.4
- LiveBench · MathematicsGPT-6 Luna by 1.987.289.1+1.9
Following instructionsGPT-5.6 Luna
- LiveBench · Instruction FollowingGPT-5.6 Luna by 4.260.155.9+4.2
Other results16 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-OmniscienceGPT-6 Luna by 11-10.30.7+11
- OSWorld 2.0GPT-6 Luna by 7.145.652.7+7.1
- HealthBench ProfessionalGPT-6 Luna by 6.255.761.9+6.2
- HealthBench Professional length-adjustedGPT-6 Luna by 5.155.760.8+5.1
- livebench_data_analysisGPT-5.6 Luna by 4.67873.4+4.6
- ARC-AGI-1GPT-5.6 Luna by 490.786.7+4
- HealthBench length-adjustedGPT-5.6 Luna by 1.355.854.5+1.3
- HealthBenchGPT-5.6 Luna by 1.255.854.6+1.2
- livebench_languageGPT-6 Luna by 1.272.673.8+1.2
- Terminal-Bench 4.0tie17.316.4tie
- AA Intelligencetie37.338.1tie
- HealthBench Consensus length-adjustedtie95.195.9tie
- Agents' Last Examtie50.350.9tie
- HealthBench Hard length-adjustedtie3231.4tie
- DeepSWE 1.1tie6766.6tie
- GDP (Surge AI)tie22.722.8tie
Questions people ask
Which is better, GPT-5.6 Luna or GPT-6 Luna?
GPT-5.6 Luna wins three of the eight areas we test: reasoning, math and following instructions. GPT-6 Luna wins coding and facts, and costs 57% less. They are level on agents, images and charts and long documents.
Which is better for coding?
GPT-6 Luna. It wins 3 of the 4 coding tests both models report; GPT-5.6 Luna wins 1.
Which is cheaper?
GPT-5.6 Luna costs $0.20 per million input tokens and $1.20 per million output tokens; GPT-6 Luna costs $0.10 and $0.50. That makes GPT-6 Luna about 57% cheaper for the same work.
How do you compare the two?
We use the 37 benchmark tests both models have published scores on. The verdict counts the 21 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 16 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.