SShogo Bench

GPT-5.4

Agent-ready GPT for coding and computer-use workflows at a lower cost

Shogo Agent Index

Supported#33 of 92
51.6range 47.2–56.1 · 64% of weight measured
Agentic coding
57
74% measured
Long-horizon autonomy
61
100% measured
Knowledge-work agents
40
52% measured
Computer use
42
estimated
Security & context learning
54
50% measured

Price

Input
$2.50/M
Output
$15.00/M
Cache read
$0.25/M
3:1 blend
$5.63/M
Agent-effective
$1.27/M

Cheapest host for agents: OpenAI at $0.64/M.

Agent-effective = 20 input tokens per output token, 85% cache hits. Why.

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verified76.9%8%source
Terminal-Bench 2.081.8%8%ForgeCode
DeepSWE51.8%6%mini-swe-agent · $5.65/task
FrontierSWEest.5%predicted 23
CursorBenchest.4%predicted 41
GSO-Bench31.4%4%OpenHands
METR Time Horizons74.3%10%source
Vending-Bench 2$6,1446%source
PostTrainBench17.8%4%Codex CLI
APEX-Agents52.4%8%source
GDPvalest.7%predicted 53
Remote Labor Indexest.5%predicted 6
DeepResearch Bench35.1%5%source
OSWorld 2.0est.6%predicted 14
OSWorldest.4%predicted 83
Cybenchest.5%predicted 81
CL-bench27.9%5%source

Other benchmarks

BenchmarkCategoryScoreSource
AlgoTuneCoding1.85×source
ARC-AGIReasoning93.7%source
ARC-AGI-2Reasoning74.0%source
Chess PuzzlesReasoning48.0%source
CritPtReasoning23.4%source
DTBenchKnowledge94.4%source
FrontierMath Tiers 1-3Math78.6%source
FrontierMath Tier 4Math49.0%source
GPQA DiamondReasoning93.3%source
Humanity's Last ExamReasoning36.2%source
LMCAKnowledge52.0%source
MirrorCodeCoding23.3%source
OTIS Mock AIMEMath97.8%source
ProofBenchMath56.0%source
SciCodeCoding56.6%source
SimpleQA VerifiedKnowledge45.1%source
WebDev ArenaCoding1463source

API hosts

Live prices and uptime for every host serving GPT-5.4 via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
OpenAI$1.25$7.50$0.13$0.64100.0%—
Azure$2.50$15.00$0.25$1.27100.0%—
OpenAI$2.50$15.00$0.25$1.27100.0%—
OpenAI$5.00$30.00$0.50$2.55100.0%—