SShogo Bench

GPT-5.5

Default frontier GPT for coding, computer use, research, and knowledge work

Shogo Agent Index

Supported#19 of 92
56.7range 52.5–61 · 60% of weight measured
Agentic coding
63
74% measured
Long-horizon autonomy
67
50% measured
Knowledge-work agents
46
72% measured
Computer use
43
60% measured
Security & context learning
55
estimated

Price

Input
$5.00/M
Output
$30.00/M
Cache read
$0.50/M
3:1 blend
$11.25/M
Agent-effective
$2.55/M

Cheapest host for agents: OpenAI at $1.27/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verified80.6%8%source
Terminal-Bench 2.084.7%8%NexAU-AHE
DeepSWE67.0%6%mini-swe-agent · $7.23/task
FrontierSWEest.5%predicted 30
CursorBenchest.4%predicted 44
GSO-Bench40.2%4%OpenHands
METR Time Horizonsest.10%predicted 80
Vending-Bench 2$7,5246%source
PostTrainBench23.9%4%Codex CLI
APEX-Agents55.1%8%source
GDPvalest.7%predicted 57
Remote Labor Index6.3%5%source
DeepResearch Bench54.0%5%source
OSWorld 2.013.0%6%source
OSWorldest.4%predicted 87
Cybenchest.5%predicted 85
CL-benchest.5%predicted 25

Other benchmarks

BenchmarkCategoryScoreSource
ARC-AGIReasoning95.0%source
ARC-AGI-2Reasoning85.0%source
Chess PuzzlesReasoning63.0%source
CritPtReasoning27.1%source
DTBenchKnowledge96.0%source
ExploitBenchSecurity & context47.4%source
FrontierCodeCoding43.0%codex
FrontierMath Tiers 1-3Math85.3%source
FrontierMath Tier 4Math72.5%source
GPQA DiamondReasoning94.0%source
LMCAKnowledge54.3%source
MirrorCodeCoding23.3%source
OTIS Mock AIMEMath100.0%source
ProofBenchMath50.0%source
SciCodeCoding56.1%source
SimpleBenchReasoning69.0%source
SimpleQA VerifiedKnowledge63.0%source
WebDev ArenaCoding1510source

API hosts

Live prices and uptime for every host serving GPT-5.5 via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
OpenAI$2.50$15.00$0.25$1.27100.0%—
Azure$5.00$30.00$0.50$2.55100.0%—
OpenAI$5.00$30.00$0.50$2.55100.0%—
OpenAI$12.50$75.00$1.25$6.37100.0%—