Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

GPT-6 Luna vs GPT-6 Sol

44 SHARED BENCHMARKS

Across 44 shared benchmarks, GPT-6 Luna scores higher on 6 and GPT-6 Sol on 36, with 2 level. The widest gap is Terminal-Bench 4.0, where GPT-6 Sol scores 43.9 against 12.6. GPT-6 Luna is the cheaper of the two on tracked API pricing ($0.10 against $2.00 per million input tokens).

OPENAIVSOPENAI44 SHARED6–36 HEAD-TO-HEADNEWEST SCORE ADDED

At a glance

MakerOpenAIOpenAI
Released
Price per 1M tokens input / output$0.10 / $0.50$2.00 / $10.00
Cost of 1M in + 1M out$0.60 20.0× less$12.00
Head-to-head of 44 shared benchmarks6 wins36 wins
Scores tracked independently verified121 33 ◆126 37 ◆

Release dates per Artificial Analysis. Prices: Artificial Analysis for GPT-6 Luna; Artificial Analysis for GPT-6 Sol. ◆ = independently verified score.

Where each leadsby capability area · benchmark wins

AreaGPT-6 LunaWINSGPT-6 Sol
Reasoning05GPT-6 Sol leads 5 of 5 · widest: CritPt 19.4 vs 30.9
Coding04GPT-6 Sol leads 4 of 4 · widest: LMArena · WebDev 1593 vs 1681
Agentic02GPT-6 Sol leads 2 of 2 · widest: Terminal-Bench 4.0 12.6 vs 43.9
Factuality02GPT-6 Sol leads 2 of 2 · widest: AA-Omniscience · Non-hallucination 23.3 vs 39.9
Instruction Following01GPT-6 Sol leads 1 of 1 · widest: LiveBench · Instruction Following 55.9 vs 68.6
Long Context01GPT-6 Sol leads 1 of 1
Math01GPT-6 Sol leads 1 of 1 · widest: LiveBench · Mathematics 89.1 vs 96.4
Multimodal01GPT-6 Sol leads 1 of 1 · widest: MMMU-Pro 75.5 vs 83.3

Biggest gaps

GPT-6 Luna pulls furthest ahead on

No ratified-area lead of 3 points or more.

GPT-6 Sol pulls furthest ahead on

  1. Terminal-Bench 4.043.9 vs 12.6
  2. AA-Omniscience · Non-hallucination39.9 vs 23.3
  3. CritPt30.9 vs 19.4

Every shared benchmark44 · grouped by area

Reasoning 5

BenchmarkGPT-6 LunaMARGINGPT-6 Sol
ARC-AGI-30.1 ◆◆ 4.6
ARC-AGI-259.3 ◆◆ 89.6
CritPt19.430.9
LiveBench · Reasoning81.8 ◆◆ 88.7

Coding 4

BenchmarkGPT-6 LunaMARGINGPT-6 Sol
LMArena · WebDev1593 ◆◆ 1681
LiveBench · Coding79 ◆◆ 81.8
SciCode54.657.6

Agentic 2

BenchmarkGPT-6 LunaMARGINGPT-6 Sol
GDPVal43.449.3
Terminal-Bench 4.012.643.9

Instruction Following 1

Long Context 1

BenchmarkGPT-6 LunaMARGINGPT-6 Sol
AA-LCR83.383.7

Math 1

BenchmarkGPT-6 LunaMARGINGPT-6 Sol
LiveBench · Mathematics89.1 ◆◆ 96.4

Multimodal 1

BenchmarkGPT-6 LunaMARGINGPT-6 Sol
MMMU-Pro75.583.3

Other shared benchmarks 27

BenchmarkGPT-6 LunaMARGINGPT-6 Sol
Agentic Safe Completions - Chat prod - Chat Plugins0.80.9
ARC-AGI-186.7 ◆◆ 95.5
DeepSWE 1.166.668.8
Dynamic Benchmarks - Emotional reliance11
Dynamic Benchmarks - Mental health11
Dynamic Benchmarks - Self-harm0.91
Image input evaluations - extremism11
Image input evaluations - harms-erotic11
Image input evaluations - hate11
Image input evaluations - self-harm11
livebench_data_analysis73.4 ◆◆ 81.2
livebench_language73.8 ◆◆ 85.3
OSWorld 2.052.764.4
Production Benchmarks - Gore0.90.9
U18 evaluations - Age-restricted goods, services, and dangerous challenges / activities0.80.9
U18 evaluations - Eating Disorders0.90.9
U18 evaluations - Emotional Reliance0.90.9
U18 evaluations - Gore0.90.9
U18 evaluations - Self Harm11
U18 evaluations - Sexual Content0.91

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.