Skip to main content

GPT-5.4 vs GPT-6 Luna

SpecGPT-5.4GPT-6 Luna
Respan Index49.9 (#44)48.6 (#46)
Blended price per 1M$5.63$0.2
Context window1.1M1.1M
Input / output per 1M$2.5 / $15$0.1 / $0.5
Max output128K128K
Inputtext, visiontext, vision
Released2026-03-052026-09-22

Respan Index by category

Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.

CategoryGPT-5.4GPT-6 Luna
Coding25.9 GPT-5.428.4 GPT-6 Luna
Agents61.9 GPT-5.460.0 GPT-6 Luna
Reasoning61.4 GPT-5.457.0 GPT-6 Luna
Knowledge & language46.9 GPT-5.438.9 GPT-6 Luna
Human preference52.9 GPT-5.455.4 GPT-6 Luna

How the Respan Index works

Benchmarks both report

GPT-5.4 leads on 14 of the 22 benchmarks both report. Higher is better. Each score links to where it was published.

BenchmarkGPT-5.4GPT-6 Luna
LMArena Elo14751443
LMArena Vision (LMArena)12821234
LMArena WebDev (LMArena)14621579
MirrorCode (Epoch AI)23.3%26.7%
Chess Puzzles (Epoch AI)48%36%
LMArena Agent (LMArena)0.02670.0064
APEX-Agents (Mercor)52.4%44.3%
FrontierMath Tier 4 v2 (Epoch AI)49%56.1%
FrontierMath Tiers 1-3 v2 (Epoch AI)78.6%78.9%
Furniture Assembly (Epoch AI)37.5%44.2%
GPQA Diamond92.8%90.5%
LiveBench Agentic Coding (LiveBench)53.8%51.2%
LiveBench Coding (LiveBench)77.5%79%
LiveBench Data Analysis (LiveBench)79.3%73.4%
LiveBench Instruction Following (LiveBench)70.2%55.9%
LiveBench Language (LiveBench)82.6%73.8%
LiveBench Mathematics (LiveBench)94.1%89.1%
LiveBench Reasoning (LiveBench)88.1%81.8%
Mystery Game Puzzles (Epoch AI)37%7%
OTIS Mock AIME 2024-2025 (Epoch AI)97.8%98.9%
SAGE (Vals AI)43.3%48.1%
SimpleQA Verified45.1%41.4%

FAQ

Is GPT-5.4 better than GPT-6 Luna?

GPT-5.4 scores 49.9 on the Respan Index (#44), ahead of GPT-6 Luna at 48.6 (#46). GPT-5.4 leads on 14 of the 22 benchmarks both report.

Which is cheaper, GPT-5.4 or GPT-6 Luna?

GPT-5.4 lists at $2.5 input and $15 output per 1M tokens; GPT-6 Luna at $0.1 and $0.5. GPT-6 Luna is 28x cheaper per token at list price.

Which has the larger context window?

Both accept 1.1M tokens of context.

Compare other models

vs

Built for AI agents.
Break less.
Ship more.