Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

Claude Opus 4.5 vs GPT-5.6 Luna

36 SHARED BENCHMARKS

Across 36 shared benchmarks, Claude Opus 4.5 scores higher on 9 and GPT-5.6 Luna on 27. The widest gap is AA-Omniscience · Non-hallucination, where Claude Opus 4.5 scores 39 against 7.4. GPT-5.6 Luna is the cheaper of the two on tracked API pricing ($0.20 against $5.00 per million input tokens).

ANTHROPICVSOPENAI36 SHARED9–27 HEAD-TO-HEADNEWEST SCORE ADDED

At a glance

MakerAnthropicOpenAI
Released
Price per 1M tokens input / output$5.00 / $25.00$0.20 / $1.20
Cost of 1M in + 1M out$30.00$1.40 21.4× less
Head-to-head of 36 shared benchmarks9 wins27 wins
Scores tracked independently verified204 44 ◆201 43 ◆

Release dates per Artificial Analysis. Prices: Anthropic's own price page for Claude Opus 4.5; Artificial Analysis for GPT-5.6 Luna. ◆ = independently verified score.

Where each leadsby capability area · benchmark wins

AreaClaude Opus 4.5WINSGPT-5.6 Luna
Reasoning05GPT-5.6 Luna leads 5 of 5 · widest: CritPt 4.6 vs 20.6
Coding06GPT-5.6 Luna leads 6 of 6 · widest: LiveBench · Agentic Coding 39.7 vs 48.4
Agentic12GPT-5.6 Luna leads 2 of 3 · widest: BrowseComp 37 vs 83.3
Factuality30Claude Opus 4.5 leads 3 of 3 · widest: AA-Omniscience · Non-hallucination 39 vs 7.4
Instruction Following10Claude Opus 4.5 leads 1 of 1
Long Context01GPT-5.6 Luna leads 1 of 1 · widest: AA-LCR 77.3 vs 83.7
Math14GPT-5.6 Luna leads 4 of 5 · widest: FrontierMath Tiers 1-3 (v2) 34.4 vs 82.1
Multimodal02GPT-5.6 Luna leads 2 of 2 · widest: CharXiv (RQ) 67.2 vs 82.7

Biggest gaps

Claude Opus 4.5 pulls furthest ahead on

  1. AA-Omniscience · Non-hallucination39 vs 7.4
  2. SimpleQA Verified45.7 vs 41
  3. AA-Omniscience · Accuracy46.6 vs 42.7

GPT-5.6 Luna pulls furthest ahead on

  1. CritPt20.6 vs 4.6
  2. FrontierMath Tiers 1-3 (v2)82.1 vs 34.4
  3. BrowseComp83.3 vs 37

Every shared benchmark36 · grouped by area

Reasoning 5

ARC-AGI-237.6◆ 59.6
CritPt4.620.6
GPQA Diamond86.691.1
LiveBench · Reasoning80.1 ◆◆ 85.6

Coding 6

Agentic 3

Instruction Following 1

Long Context 1

AA-LCR77.383.7

Math 5

Multimodal 2

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 (anthropic-official, direct), otherwise the lowest tracked offer.

Compare Claude Opus 4.5 withALL PAIRINGS →

Compare GPT-5.6 Luna withALL PAIRINGS →