Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

GPT-5.4 vs GPT-6 Luna

40 SHARED BENCHMARKS

Across 40 shared benchmarks, GPT-5.4 scores higher on 21 and GPT-6 Luna on 18, with 1 level. The widest gap is AA-Omniscience, where GPT-5.4 scores 5.8 against 0.7. GPT-6 Luna is the cheaper of the two on tracked API pricing ($0.10 against $2.50 per million input tokens).

OPENAIVSOPENAI40 SHARED21–18 HEAD-TO-HEADNEWEST SCORE ADDED

At a glance

MakerOpenAIOpenAI
Released
Price per 1M tokens input / output$2.50 / $15.00$0.10 / $0.50
Cost of 1M in + 1M out$17.50$0.60 29.2× less
Head-to-head of 40 shared benchmarks21 wins18 wins
Scores tracked independently verified199 41 ◆121 33 ◆

Release dates per Artificial Analysis. Prices: Artificial Analysis for GPT-5.4; Artificial Analysis for GPT-6 Luna. ◆ = independently verified score.

Where each leadsby capability area · benchmark wins

AreaGPT-5.4WINSGPT-6 Luna
Reasoning50GPT-5.4 leads 5 of 5 · widest: ARC-AGI-2 73.3 vs 59.3
Coding22Even, 2–2 of 4
Agentic01GPT-6 Luna leads 1 of 1 · widest: GDPVal 36.6 vs 43.4
Factuality11Even, 1–1 of 2
Instruction Following10GPT-5.4 leads 1 of 1 · widest: LiveBench · Instruction Following 70.2 vs 55.9
Long Context01GPT-6 Luna leads 1 of 1
Math10GPT-5.4 leads 1 of 1 · widest: LiveBench · Mathematics 94.2 vs 89.1
Multimodal10GPT-5.4 leads 1 of 1

Biggest gaps

GPT-5.4 pulls furthest ahead on

  1. LiveBench · Instruction Following70.2 vs 55.9
  2. ARC-AGI-273.3 vs 59.3
  3. CritPt23.4 vs 19.4

GPT-6 Luna pulls furthest ahead on

  1. AA-Omniscience · Non-hallucination23.3 vs 8.3
  2. GDPVal43.4 vs 36.6
  3. LMArena · WebDev1593 vs 1466

Every shared benchmark40 · grouped by area

Reasoning 5

BenchmarkGPT-5.4MARGINGPT-6 Luna
ARC-AGI-30.2 ◆◆ 0.1
ARC-AGI-273.3◆ 59.3
CritPt23.419.4
LiveBench · Reasoning88.1 ◆◆ 81.8

Coding 4

BenchmarkGPT-5.4MARGINGPT-6 Luna
LMArena · WebDev1466 ◆◆ 1593
LiveBench · Coding77.5 ◆◆ 79
SciCode56.654.6

Agentic 1

BenchmarkGPT-5.4MARGINGPT-6 Luna
GDPVal36.643.4

Instruction Following 1

Long Context 1

BenchmarkGPT-5.4MARGINGPT-6 Luna
AA-LCR8283.3

Math 1

BenchmarkGPT-5.4MARGINGPT-6 Luna
LiveBench · Mathematics94.2 ◆◆ 89.1

Multimodal 1

BenchmarkGPT-5.4MARGINGPT-6 Luna
MMMU-Pro78.475.5

Other shared benchmarks 24

BenchmarkGPT-5.4MARGINGPT-6 Luna
ARC-AGI-193.7 ◆◆ 86.7
Dynamic Benchmarks - Emotional reliance11
Dynamic Benchmarks - Mental health0.91
Dynamic Benchmarks - Self-harm10.9
Image input evaluations - extremism11
Image input evaluations - harms-erotic11
Image input evaluations - hate11
Image input evaluations - self-harm11
livebench_data_analysis79.3 ◆◆ 73.4
livebench_language82.6 ◆◆ 73.8
Production Benchmarks - Gore0.80.9
U18 evaluations - Age-restricted goods, services, and dangerous challenges / activities0.80.8
U18 evaluations - Eating Disorders0.70.9
U18 evaluations - Emotional Reliance0.90.9
U18 evaluations - Gore0.80.9
U18 evaluations - Self Harm11
U18 evaluations - Sexual Content0.90.9

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 GPT-5.4 withALL PAIRINGS →