SShogo Bench

GPT-5.4 vs GPT-5.6 Luna

GPT-5.4 leads GPT-5.6 Luna by 0.2 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. GPT-5.4 wins 12 and GPT-5.6 Luna wins 6 of 18 shared benchmarks. GPT-5.6 Luna is cheaper for agent workloads.

Add more models to this comparison →

Pillars

PillarGPT-5.4GPT-5.6 Luna
Agentic coding5759
Long-horizon autonomy6161
Knowledge-work agents4039
Computer use4240
Security & context learning5451

Shared benchmarks

BenchmarkGPT-5.4GPT-5.6 Luna
APEX-Agentsindex52.4%43.0%
DeepSWEindex51.8%67.2%
Vending-Bench 2index$6,144$4,095
ARC-AGI93.7%88.0%
ARC-AGI-274.0%59.5%
Chess Puzzles48.0%43.0%
CritPt23.4%20.6%
DTBench94.4%89.1%
FrontierMath Tier 449.0%61.0%
FrontierMath Tiers 1-378.6%82.1%
GPQA Diamond93.3%91.6%
LMCA52.0%48.5%
MirrorCode23.3%21.1%
OTIS Mock AIME97.8%98.3%
ProofBench56.0%60.0%
SciCode56.6%53.6%
SimpleQA Verified45.1%41.0%
WebDev Arena14631519

Specs and price

GPT-5.4GPT-5.6 Luna
Agent Index51.6 (#33)51.4 (#34)
Agent $/M$1.27$0.10
Input / output $/M$2.50 / $15.00$0.20 / $1.20
Context1.1M1.1M
ReleasedMar 5, 2026Jul 9, 2026
WeightsClosedClosed
API hosts45

Don’t pick one. Route.

Shogo can run GPT-5.4 and GPT-5.6 Luna side by side and send each step of an agent to the one that’s cheaper for it.

Try both in Shogo