SShogo Bench

GPT-5.4 mini

Strong small GPT for coding subagents, quick tool use, and high-volume work

Shogo Agent Index

Unranked

Not ranked yet: 1 agentic benchmark measured, and we need at least 2 plus an overall capability score. It will be ranked as results are published.

Agentic coding
40
estimated
Long-horizon autonomy
46
estimated
Knowledge-work agents
28
20% measured
Computer use
26
estimated
Security & context learning
39
estimated

Price

Input
$0.75/M
Output
$4.50/M
Cache read
$0.075/M
3:1 blend
$1.69/M
Agent-effective
$0.38/M

Cheapest host for agents: OpenAI at $0.19/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 71
Terminal-Bench 2.0est.8%predicted 50
DeepSWEest.6%predicted 34
FrontierSWEest.5%predicted 7
CursorBenchest.4%predicted 31
GSO-Benchest.4%predicted 15
METR Time Horizonsest.10%predicted 69
Vending-Bench 2est.6%predicted 25
PostTrainBenchest.4%predicted 20
APEX-Agentsest.8%predicted 32
GDPvalest.7%predicted 37
Remote Labor Indexest.5%predicted 2
DeepResearch Bench36.3%5%source
OSWorld 2.0est.6%predicted 4
OSWorldest.4%predicted 59
Cybenchest.5%predicted 60
CL-benchest.5%predicted 19

Other benchmarks

BenchmarkCategoryScoreSource
ARC-AGIReasoning63.7%source
ARC-AGI-2Reasoning18.9%source
Chess PuzzlesReasoning23.0%source
CritPtReasoning10.0%source
DTBenchKnowledge80.0%source
FrontierCodeCoding27.0%codex
FrontierMath Tiers 1-3Math51.2%source
FrontierMath Tier 4Math9.8%source
GPQA DiamondReasoning86.9%source
LMCAKnowledge40.8%source
OTIS Mock AIMEMath88.9%source
ProofBenchMath21.0%source
SciCodeCoding52.1%source
SimpleQA VerifiedKnowledge29.4%source
WebDev ArenaCoding1397source

API hosts

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

HostInputOutputCache readAgent $/MUptime 24hQuant
OpenAI$0.38$2.25$0.037$0.19100.0%—
Azure$0.75$4.50$0.075$0.38100.0%—
OpenAI$0.75$4.50$0.075$0.38100.0%—
OpenAI$1.50$9.00$0.15$0.76100.0%—