GPT-5.6 Luna vs Kimi K3
Wins 1 of 8 areas
Math
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents · Following instructions
Kimi K3 is the stronger all-rounder.GPT-5.6 Luna is cheaper and better at math.
Scores updated · 48 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.
- AgentsCarrying out multi-step tasks on its own07Kimi K37 of 7 tests
- ReasoningHard problems that need careful thinking05Kimi K35 of 6 tests · 1 tie
- CodingWriting and fixing software14Kimi K34 of 5 tests
- FactsGetting facts right instead of making them up03Kimi K33 of 3 tests
- Images and chartsUnderstanding pictures, charts and video02Kimi K32 of 2 tests
- Long documentsFinding answers in very long texts01Kimi K31 of 1 test
- Following instructionsDoing exactly what it is asked01Kimi K31 of 1 test
- MathCompetition and research-level math40GPT-5.6 Luna4 of 5 tests · 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 92% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Kimi K3 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+39.4points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+14.9points ahead
- Common-sense trick questionsSimpleBench+13.9points ahead
Where GPT-5.6 Luna pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+19.5points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+9.9points ahead
- Fresh competition math problemsLiveBench · Mathematics+2.8points ahead
Every test, side by side
All 48 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- LMArena · WebDevKimi K3 by 139 rating points15191658+139 rating
- LiveBench · Agentic CodingKimi K3 by 13.848.462.2+13.8
- SciCodeKimi K3 by 5.953.659.5+5.9
- Terminal-Bench 2.1Kimi K3 by 4.180.985+4.1
- LiveBench · CodingGPT-5.6 Luna by 1.482.981.5+1.4
AgentsKimi K3
- τ-Bench V3 · BankingKimi K3 by 14.931.146+14.9
- OSWorld-VerifiedKimi K3 by 12.272.684.8+12.2
- BrowseCompKimi K3 by 7.983.391.2+7.9
- AA IT-Bench SREKimi K3 by 7.440.347.7+7.4
- AA ApexAgentsKimi K3 by 5.535.841.3+5.5
- GDPValKimi K3 by 3.74851.7+3.7
- Terminal-Bench 4.0Kimi K3 by 111.612.6+1
ReasoningKimi K3
- SimpleBenchKimi K3 by 13.946.860.7+13.9
- Humanity's Last ExamKimi K3 by 7.439.546.9+7.4
- LiveBench · ReasoningKimi K3 by 5.185.690.7+5.1
- CritPtKimi K3 by 2.820.623.4+2.8
- GPQA DiamondKimi K3 by 2.491.193.5+2.4
- ARC-AGI-2tie59.660.4tie
FactsKimi K3
- AA-Omniscience · Non-hallucinationKimi K3 by 39.47.446.8+39.4
- SimpleQA VerifiedKimi K3 by 9.64150.6+9.6
- AA-Omniscience · AccuracyKimi K3 by 4.942.747.6+4.9
Images and chartsKimi K3
- CharXiv (RQ)Kimi K3 by 8.682.791.3+8.6
- MMMU-ProKimi K3 by 1.978.680.5+1.9
MathGPT-5.6 Luna
- FrontierMath Tier 4GPT-5.6 Luna by 19.558.539+19.5
- FrontierMath Tiers 1-3 (v2)GPT-5.6 Luna by 9.982.172.2+9.9
- LiveBench · MathematicsGPT-5.6 Luna by 2.887.284.4+2.8
- HMMT Feb. 2026GPT-5.6 Luna by 1.598.597+1.5
- AIME 2026tie97.696.7tie
Following instructionsKimi K3
- LiveBench · Instruction FollowingKimi K3 by 11.360.171.4+11.3
Other results18 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-OmniscienceKimi K3 by 30-10.319.7+30
- Agents' Last ExamGPT-5.6 Luna by 2250.328.3+22
- ToolathlonKimi K3 by 19.853.473.2+19.8
- GDPval-AA v2Kimi K3 by 152 rating points15301682+152 rating
- livebench_languageKimi K3 by 12.972.685.5+12.9
- OSWorld 2.0Kimi K3 by 12.745.658.3+12.7
- HLE (with tools)Kimi K3 by 10.948.959.8+10.9
- GDPval-AA v2 EloKimi K3 by 102 rating points15841686+102 rating
- Toolathlon VerifiedKimi K3 by 8.667.976.5+8.6
- AA Agentic IndexKimi K3 by 7.942.750.6+7.9
- AA IntelligenceKimi K3 by 6.337.343.6+6.3
- Artificial Analysis Coding IndexKimi K3 by 4.871.476.2+4.8
- Terminal-Bench 4.0GPT-5.6 Luna by 4.717.312.6+4.7
- ARC-AGI-1Kimi K3 by 3.890.794.5+3.8
- DeepSWE 1.1Kimi K3 by 26769+2
- GDP (Surge AI)tie22.722tie
- livebench_data_analysistie7878.7tie
- DeepSWEtie67.267.5tie
Questions people ask
Which is better, GPT-5.6 Luna or Kimi K3?
Kimi K3 wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, long documents and following instructions. GPT-5.6 Luna wins math, and costs 92% less.
Which is better for coding?
Kimi K3. It wins 4 of the 5 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; Kimi K3 costs $3.00 and $15.00. That makes GPT-5.6 Luna about 92% cheaper for the same work.
How do you compare the two?
We use the 48 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 18 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.