Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

GPT-5.6 Luna vs Kimi K3

45 SHARED BENCHMARKS

Across 45 shared benchmarks, GPT-5.6 Luna scores higher on 7 and Kimi K3 on 38. The widest gap is AA-Omniscience · Non-hallucination, where Kimi K3 scores 46.8 against 7.4. GPT-5.6 Luna is the cheaper of the two on tracked API pricing ($0.20 against $3.00 per million input tokens).

OPENAIVSMOONSHOT45 SHARED7–38 HEAD-TO-HEADNEWEST SCORE ADDED

At a glance

MakerOpenAIMoonshot
Released
Price per 1M tokens input / output$0.20 / $1.20$3.00 / $15.00
Cost of 1M in + 1M out$1.40 12.9× less$18.00
Head-to-head of 45 shared benchmarks7 wins38 wins
Scores tracked independently verified183 43 ◆131 26 ◆

Release dates per Artificial Analysis. Prices: Artificial Analysis for GPT-5.6 Luna; Artificial Analysis for Kimi K3. ◆ = independently verified score.

Where each leadsby capability area · benchmark wins

AreaGPT-5.6 LunaWINSKimi K3
Reasoning06Kimi K3 leads 6 of 6 · widest: SimpleBench 46.8 vs 60.7
Coding14Kimi K3 leads 4 of 5 · widest: LiveBench · Agentic Coding 48.4 vs 62.2
Agentic07Kimi K3 leads 7 of 7 · widest: τ-Bench V3 · Banking 31.1 vs 46
Factuality03Kimi K3 leads 3 of 3 · widest: AA-Omniscience · Non-hallucination 7.4 vs 46.8
Instruction Following01Kimi K3 leads 1 of 1 · widest: LiveBench · Instruction Following 60.1 vs 71.4
Long Context01Kimi K3 leads 1 of 1 · widest: AA-LCR 83.7 vs 88.7
Math50GPT-5.6 Luna leads 5 of 5 · widest: FrontierMath Tier 4 58.5 vs 39
Multimodal02Kimi K3 leads 2 of 2 · widest: CharXiv (RQ) 82.7 vs 91.3

Biggest gaps

GPT-5.6 Luna pulls furthest ahead on

  1. FrontierMath Tier 458.5 vs 39
  2. FrontierMath Tiers 1-3 (v2)82.1 vs 72.2

Kimi K3 pulls furthest ahead on

  1. AA-Omniscience · Non-hallucination46.8 vs 7.4
  2. τ-Bench V3 · Banking46 vs 31.1
  3. SimpleBench60.7 vs 46.8

Every shared benchmark45 · grouped by area

Reasoning 6

BenchmarkGPT-5.6 LunaMARGINKimi K3
ARC-AGI-259.6 ◆◆ 60.4
CritPt20.623.4
GPQA Diamond91.193.5
LiveBench · Reasoning85.6 ◆◆ 90.7
SimpleBench46.8 ◆◆ 60.7

Coding 5

BenchmarkGPT-5.6 LunaMARGINKimi K3
LMArena · WebDev1520 ◆◆ 1660
LiveBench · Coding82.9 ◆◆ 81.5
SciCode53.659.5

Agentic 7

BenchmarkGPT-5.6 LunaMARGINKimi K3
BrowseComp83.391.2
GDPVal47.251.2
Terminal-Bench 4.011.612.6

Instruction Following 1

Long Context 1

BenchmarkGPT-5.6 LunaMARGINKimi K3
AA-LCR83.788.7

Math 5

BenchmarkGPT-5.6 LunaMARGINKimi K3
AIME 202697.6◆ 96.7
HMMT Feb. 202698.5◆ 97
LiveBench · Mathematics87.2 ◆◆ 84.4

Multimodal 2

BenchmarkGPT-5.6 LunaMARGINKimi K3
CharXiv (RQ)82.791.3
MMMU-Pro78.680.5

Other shared benchmarks 15

Each score is the one the model's own page shows — the most authoritative tracked result for that benchmark, on the benchmark's standard methodology (source-attributed). ↓ marks lower-is-better metrics; ◆ an independently verified score. Open either model for its full surface, provenance and pricing. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (direct, direct), otherwise the lowest tracked offer.

Compare Kimi K3 withALL PAIRINGS →