GPT-5.5 vs GPT-5.6 Luna
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Following instructions
Wins 0 of 8 areas
—
GPT-5.5 is the stronger all-rounder.GPT-5.6 Luna is cheaper.
Scores updated · 61 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 thinking60GPT-5.56 of 7 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own61GPT-5.56 of 7 tests
- MathCompetition and research-level math40GPT-5.54 of 5 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video40GPT-5.54 of 4 tests
- FactsGetting facts right instead of making them up30GPT-5.53 of 3 tests
- CodingWriting and fixing software32GPT-5.53 of 6 tests · 1 tie
- Following instructionsDoing exactly what it is asked10GPT-5.51 of 1 test
- 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-5.6 Luna costs 96% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where GPT-5.5 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+25.4points ahead
- Common-sense trick questionsSimpleBench+22.2points ahead
- Short factual questions, answered correctlySimpleQA Verified+22points ahead
Where GPT-5.6 Luna pulls ahead
- Real work tasks from 44 professionsGDPVal+5.5points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+4.1points ahead
Every test, side by side
All 61 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.5
- LiveBench · Agentic CodingGPT-5.5 by 5.65448.4+5.6
- SWE-bench ProGPT-5.6 Luna by 4.158.662.7+4.1
- Terminal-Bench 2.1GPT-5.5 by 3.484.380.9+3.4
- LMArena · WebDevGPT-5.6 Luna by 32 rating points14871519+32 rating
- SciCodeGPT-5.5 by 2.255.853.6+2.2
- LiveBench · Codingtie82.282.9tie
AgentsGPT-5.5
- τ-Bench V3 · BankingGPT-5.5 by 7.93931.1+7.9
- OSWorld-VerifiedGPT-5.5 by 6.178.772.6+6.1
- AA IT-Bench SREGPT-5.5 by 5.545.840.3+5.5
- GDPValGPT-5.6 Luna by 5.542.548+5.5
- Terminal-Bench 4.0GPT-5.5 by 314.611.6+3
- AA ApexAgentsGPT-5.5 by 1.937.735.8+1.9
- BrowseCompGPT-5.5 by 1.184.483.3+1.1
ReasoningGPT-5.5
- ARC-AGI-2GPT-5.5 by 25.48559.6+25.4
- SimpleBenchGPT-5.5 by 22.26946.8+22.2
- CritPtGPT-5.5 by 6.527.120.6+6.5
- Humanity's Last ExamGPT-5.5 by 6.345.839.5+6.3
- LiveBench · ReasoningGPT-5.5 by 4.189.785.6+4.1
- GPQA DiamondGPT-5.5 by 2.493.591.1+2.4
- ARC-AGI-3tie0.40.2tie
FactsGPT-5.5
- SimpleQA VerifiedGPT-5.5 by 226341+22
- AA-Omniscience · AccuracyGPT-5.5 by 15.35842.7+15.3
- AA-Omniscience · Non-hallucinationGPT-5.5 by 3.6117.4+3.6
Images and chartsGPT-5.5
- BLINKGPT-5.5 by 9.778.368.6+9.7
- MathVistaGPT-5.5 by 7.384.276.9+7.3
- LMArena · VisionGPT-5.5 by 39 rating points12951256+39 rating
- MMMU-ProGPT-5.5 by 1.379.978.6+1.3
MathGPT-5.5
- FrontierMath Tier 4GPT-5.5 by 1472.558.5+14
- LiveBench · MathematicsGPT-5.5 by 8.795.987.2+8.7
- FrontierMath Tiers 1-3 (v2)GPT-5.5 by 3.285.382.1+3.2
- AIME 2026GPT-5.5 by 2.410097.6+2.4
- HMMT Feb. 2026tie98.598.5tie
Following instructionsGPT-5.5
- LiveBench · Instruction FollowingGPT-5.5 by 10.670.760.1+10.6
Other results27 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 43.235.478.6+43.2
- AA-OmniscienceGPT-5.5 by 30.820.5-10.3+30.8
- Agents' Last ExamGPT-5.6 Luna by 23.726.650.3+23.7
- livebench_languageGPT-5.5 by 14.887.472.6+14.8
- RealWorldQAGPT-5.5 by 9.882.272.4+9.8
- GDPval-AA v2 EloGPT-5.6 Luna by 93 rating points14911584+93 rating
- ERQAGPT-5.5 by 864.556.5+8
- EmbSpatial-BenchGPT-5.5 by 6.481.975.5+6.4
- Toolathlon VerifiedGPT-5.5 by 5.673.567.9+5.6
- AA Agentic IndexGPT-5.6 Luna by 5.437.342.7+5.4
- SimpleVQAGPT-5.5 by 5.458.653.2+5.4
- ARC-AGI-1GPT-5.5 by 4.39590.7+4.3
- HealthBench ProfessionalGPT-5.6 Luna by 3.951.855.7+3.9
- HealthBench Professional length-adjustedGPT-5.6 Luna by 3.951.855.7+3.9
- livebench_data_analysisGPT-5.5 by 3.681.678+3.6
- Artificial Analysis Coding IndexGPT-5.5 by 3.574.971.4+3.5
- HLE (with tools)GPT-5.5 by 3.352.248.9+3.3
- DeepSWEGPT-5.5 by 2.87067.2+2.8
- GDP (Surge AI)GPT-5.5 by 2.224.922.7+2.2
- ToolathlonGPT-5.5 by 2.255.653.4+2.2
- GDPval-AA v2GPT-5.6 Luna by 21 rating points15091530+21 rating
- AA IntelligenceGPT-5.5 by 1.138.437.3+1.1
- HealthBenchtie56.555.8tie
- HealthBench length-adjustedtie56.555.8tie
- HealthBench Consensus length-adjustedtie95.695.1tie
- HealthBench Hard length-adjustedtie31.532tie
- DeepSWE 1.1tie6767tie
Questions people ask
Which is better, GPT-5.5 or GPT-5.6 Luna?
GPT-5.5 wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, math and following instructions. GPT-5.6 Luna wins none, but costs 96% less. They are level on long documents.
Which is better for coding?
GPT-5.5. It wins 3 of the 6 coding tests both models report; GPT-5.6 Luna wins 2, and 1 is a tie.
Which is cheaper?
GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens; GPT-5.6 Luna costs $0.20 and $1.20. That makes GPT-5.6 Luna about 96% cheaper for the same work.
How do you compare the two?
We use the 61 benchmark tests both models have published scores on. The verdict counts the 34 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 27 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.