Gemini 3.1 Pro vs GPT-5.6 Luna
Wins 4 of 8 areas
Reasoning · Facts · Images and charts · Following instructions
Wins 4 of 8 areas
Coding · Agents · Math · Long documents
The two are evenly matched.Gemini 3.1 Pro is better at images and charts and facts; GPT-5.6 Luna at coding and agents.
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.
- CodingWriting and fixing software16GPT-5.6 Luna6 of 7 tests
- AgentsCarrying out multi-step tasks on its own25GPT-5.6 Luna5 of 7 tests
- MathCompetition and research-level math13GPT-5.6 Luna3 of 5 tests · 1 tie
- Long documentsFinding answers in very long texts02GPT-5.6 Luna2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video50Gemini 3.1 Pro5 of 6 tests · 1 tie
- FactsGetting facts right instead of making them up30Gemini 3.1 Pro3 of 3 tests
- ReasoningHard problems that need careful thinking42Gemini 3.1 Pro4 of 7 tests · 1 tie
- Following instructionsDoing exactly what it is asked10Gemini 3.1 Pro1 of 1 test
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 90% 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
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+48.5points ahead
- Real work tasks from 44 professionsGDPVal+33.4points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+31.7points ahead
Where Gemini 3.1 Pro pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+41.7points ahead
- Common-sense trick questionsSimpleBench+32.8points ahead
- Short factual questions, answered correctlySimpleQA Verified+32.5points 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-5.6 Luna
- SWE-bench VerifiedGPT-5.6 Luna by 12.480.693+12.4
- SWE-bench ProGPT-5.6 Luna by 8.554.262.7+8.5
- LMArena · WebDevGPT-5.6 Luna by 73 rating points14461519+73 rating
- Terminal-Bench 2.1GPT-5.6 Luna by 7.173.880.9+7.1
- LiveBench · CodingGPT-5.6 Luna by 6.476.582.9+6.4
- SciCodeGemini 3.1 Pro by 5.158.753.6+5.1
- LiveBench · Agentic CodingGPT-5.6 Luna by 4.344.148.4+4.3
AgentsGPT-5.6 Luna
- GDPValGPT-5.6 Luna by 33.414.648+33.4
- AA IT-Bench SREGPT-5.6 Luna by 1030.340.3+10
- τ-Bench V3 · BankingGPT-5.6 Luna by 9.721.431.1+9.7
- Terminal-Bench 4.0GPT-5.6 Luna by 7.6411.6+7.6
- AA ApexAgentsGPT-5.6 Luna by 3.83235.8+3.8
- OSWorld-VerifiedGemini 3.1 Pro by 3.676.272.6+3.6
- BrowseCompGemini 3.1 Pro by 2.685.983.3+2.6
ReasoningGemini 3.1 Pro
- SimpleBenchGemini 3.1 Pro by 32.879.646.8+32.8
- ARC-AGI-2Gemini 3.1 Pro by 17.577.159.6+17.5
- Humanity's Last ExamGemini 3.1 Pro by 7.54739.5+7.5
- GPQA DiamondGemini 3.1 Pro by 394.191.1+3
- CritPtGPT-5.6 Luna by 2.917.720.6+2.9
- LiveBench · ReasoningGPT-5.6 Luna by 1.68485.6+1.6
- ARC-AGI-3tie0.40.2tie
FactsGemini 3.1 Pro
- AA-Omniscience · Non-hallucinationGemini 3.1 Pro by 41.749.17.4+41.7
- SimpleQA VerifiedGemini 3.1 Pro by 32.573.541+32.5
- AA-Omniscience · AccuracyGemini 3.1 Pro by 12.254.942.7+12.2
Images and chartsGemini 3.1 Pro
- MathVistaGemini 3.1 Pro by 13.390.276.9+13.3
- BLINKGemini 3.1 Pro by 10.579.168.6+10.5
- Video-MMEGemini 3.1 Pro by 5.186.781.6+5.1
- LMArena · VisionGemini 3.1 Pro by 40 rating points12961256+40 rating
- MMMU-ProGemini 3.1 Pro by 3.882.478.6+3.8
- CharXiv (RQ)tie83.382.7tie
MathGPT-5.6 Luna
- FrontierMath Tier 4GPT-5.6 Luna by 31.726.858.5+31.7
- FrontierMath Tiers 1-3 (v2)GPT-5.6 Luna by 22.459.682.1+22.4
- HMMT Feb. 2026GPT-5.6 Luna by 3.894.798.5+3.8
- LiveBench · MathematicsGemini 3.1 Pro by 3.89187.2+3.8
- AIME 2026tie98.397.6tie
Long documentsGPT-5.6 Luna
- MRCR v2 8 needle 128k (average)GPT-5.6 Luna by 48.526.374.8+48.5
- AA-LCRGPT-5.6 Luna by 1.78283.7+1.7
Following instructionsGemini 3.1 Pro
- LiveBench · Instruction FollowingGemini 3.1 Pro by 1979.160.1+19
Other results23 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.
- GDPval-AA v2 EloGPT-5.6 Luna by 619 rating points9651584+619 rating
- DeepSWEGPT-5.6 Luna by 57.21067.2+57.2
- GDPval-AA v2GPT-5.6 Luna by 568 rating points9621530+568 rating
- DeepSWE 1.1GPT-5.6 Luna by 551267+55
- AA-OmniscienceGemini 3.1 Pro by 42.231.9-10.3+42.2
- AA Agentic IndexGPT-5.6 Luna by 32.410.342.7+32.4
- SimpleVQAGemini 3.1 Pro by 16.769.953.2+16.7
- ERQAGemini 3.1 Pro by 14.370.856.5+14.3
- RealWorldQAGemini 3.1 Pro by 1385.472.4+13
- livebench_languageGemini 3.1 Pro by 12.885.472.6+12.8
- EmbSpatial-BenchGemini 3.1 Pro by 8.784.275.5+8.7
- SWEBench Pro PublicGPT-5.6 Luna by 8.554.262.7+8.5
- τ³-Bench BankingGPT-5.6 Luna by 7.816.524.3+7.8
- AA IntelligenceGPT-5.6 Luna by 7.629.737.3+7.6
- ARC-AGI-1Gemini 3.1 Pro by 7.39890.7+7.3
- LVBenchGemini 3.1 Pro by 7.366.258.9+7.3
- MLE-BenchGPT-5.6 Luna by 542.647.6+5
- GDP (Surge AI)GPT-5.6 Luna by 4.917.822.7+4.9
- ToolathlonGPT-5.6 Luna by 4.648.853.4+4.6
- Artificial Analysis Coding IndexGPT-5.6 Luna by 2.668.871.4+2.6
- HLE (with tools)Gemini 3.1 Pro by 2.551.448.9+2.5
- livebench_data_analysistie78.578tie
- HealthBench Hardtie31.832tie
Questions people ask
Which is better, Gemini 3.1 Pro or GPT-5.6 Luna?
Gemini 3.1 Pro and GPT-5.6 Luna each win four of the eight areas we test. Gemini 3.1 Pro wins reasoning, facts, images and charts and following instructions; GPT-5.6 Luna wins coding, agents, math and long documents. GPT-5.6 Luna costs 90% less.
Which is better for coding?
GPT-5.6 Luna. It wins 6 of the 7 coding tests both models report; Gemini 3.1 Pro wins 1.
Which is cheaper?
Gemini 3.1 Pro costs $2.00 per million input tokens and $12.00 per million output tokens; GPT-5.6 Luna costs $0.20 and $1.20. That makes GPT-5.6 Luna about 90% 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 38 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 23 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.