GPT-5.2 vs GPT-5.6 Luna
Wins 2 of 8 areas
Facts · Following instructions
Wins 4 of 8 areas
Coding · Agents · Reasoning · Math
GPT-5.6 Luna wins more areas, narrowly.GPT-5.2 is better at facts.
Scores updated · 49 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 thinking05GPT-5.6 Luna5 of 6 tests · 1 tie
- CodingWriting and fixing software15GPT-5.6 Luna5 of 6 tests
- AgentsCarrying out multi-step tasks on its own03GPT-5.6 Luna3 of 4 tests · 1 tie
- MathCompetition and research-level math13GPT-5.6 Luna3 of 5 tests · 1 tie
- FactsGetting facts right instead of making them up21GPT-5.22 of 3 tests
- Following instructionsDoing exactly what it is asked10GPT-5.21 of 1 test
- Images and chartsUnderstanding pictures, charts and video22Even2 each · 1 tie
- Long documentsFinding answers in very long texts11Even1 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 91% 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
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+32.8points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+26.8points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+20points ahead
Where GPT-5.2 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+11.4points ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+9points ahead
- Fresh competition math problemsLiveBench · Mathematics+6points ahead
Every test, side by side
All 49 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.6 Luna
- SWE-bench ProGPT-5.6 Luna by 32.829.962.7+32.8
- SWE-bench VerifiedGPT-5.6 Luna by 138093+13
- LMArena · WebDevGPT-5.6 Luna by 98 rating points14211519+98 rating
- LiveBench · CodingGPT-5.6 Luna by 6.876.182.9+6.8
- LiveBench · Agentic CodingGPT-5.2 by 1.950.348.4+1.9
- SciCodeGPT-5.6 Luna by 1.552.153.6+1.5
AgentsGPT-5.6 Luna
- τ-Bench V3 · BankingGPT-5.6 Luna by 2011.131.1+20
- BrowseCompGPT-5.6 Luna by 17.565.883.3+17.5
- AA ApexAgentsGPT-5.6 Luna by 12.82335.8+12.8
- GDPValtie48.348tie
ReasoningGPT-5.6 Luna
- CritPtGPT-5.6 Luna by 911.620.6+9
- ARC-AGI-2GPT-5.6 Luna by 6.752.959.6+6.7
- LiveBench · ReasoningGPT-5.6 Luna by 2.483.285.6+2.4
- Humanity's Last ExamGPT-5.6 Luna by 1.837.739.5+1.8
- SimpleBenchGPT-5.6 Luna by 145.846.8+1
- GPQA Diamondtie90.391.1tie
FactsGPT-5.2
- AA-Omniscience · Non-hallucinationGPT-5.2 by 11.418.87.4+11.4
- SimpleQA VerifiedGPT-5.6 Luna by 3.937.141+3.9
- AA-Omniscience · AccuracyGPT-5.2 by 1.644.342.7+1.6
Images and chartsEven
- MathVistaGPT-5.2 by 5.982.876.9+5.9
- Video-MMEGPT-5.2 by 4.48681.6+4.4
- MMMU-ProGPT-5.6 Luna by 474.678.6+4
- LMArena · VisionGPT-5.6 Luna by 11 rating points12451256+11 rating
- CharXiv (RQ)tie82.182.7tie
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 14.767.482.1+14.7
- LiveBench · MathematicsGPT-5.2 by 693.287.2+6
- HMMT Feb. 2026GPT-5.6 Luna by 1.59798.5+1.5
- AIME 2026tie98.397.6tie
Long documentsEven
- MRCR v2 8 needle 128k (average)GPT-5.2 by 983.874.8+9
- AA-LCRGPT-5.6 Luna by 182.783.7+1
Following instructionsGPT-5.2
- LiveBench · Instruction FollowingGPT-5.2 by 1.761.860.1+1.7
Other results17 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 38.340.378.6+38.3
- Artificial Analysis Coding IndexGPT-5.6 Luna by 22.748.771.4+22.7
- AA Agentic IndexGPT-5.2 by 17.560.242.7+17.5
- BrowseComp (context management)GPT-5.6 Luna by 147084+14
- HealthBench Professional length-adjustedGPT-5.6 Luna by 9.845.955.7+9.8
- AA-OmniscienceGPT-5.2 by 9.4-0.9-10.3+9.4
- livebench_languageGPT-5.2 by 7.279.872.6+7.2
- ToolathlonGPT-5.6 Luna by 7.146.353.4+7.1
- AA IntelligenceGPT-5.6 Luna by 6.930.437.3+6.9
- VideoMMMUGPT-5.2 by 4.885.981.1+4.8
- ARC-AGI-1GPT-5.6 Luna by 4.586.290.7+4.5
- HLE (with tools)GPT-5.6 Luna by 3.445.548.9+3.4
- SimpleVQAGPT-5.2 by 2.655.853.2+2.6
- HealthBench Hard length-adjustedGPT-5.2 by 2.334.332+2.3
- HealthBench length-adjustedGPT-5.2 by 156.855.8+1
- HealthBench Consensus length-adjustedtie94.495.1tie
- livebench_data_analysistie78.278tie
Questions people ask
Which is better, GPT-5.2 or GPT-5.6 Luna?
GPT-5.6 Luna wins four of the eight areas we test: coding, agents, reasoning and math. GPT-5.2 wins facts and following instructions. They are level on images and charts and long documents.
Which is better for coding?
GPT-5.6 Luna. It wins 5 of the 6 coding tests both models report; GPT-5.2 wins 1.
Which is cheaper?
GPT-5.2 costs $1.75 per million input tokens and $14.00 per million output tokens; GPT-5.6 Luna costs $0.20 and $1.20. That makes GPT-5.6 Luna about 91% cheaper for the same work.
How do you compare the two?
We use the 49 benchmark tests both models have published scores on. The verdict counts the 32 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 17 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.