SShogo Bench

GPT-5.1

Sharper GPT-5 generation for coding, product work, and tool-assisted tasks

Shogo Agent Index

Supported#57 of 92
38.8range 30.4–47.3 · 36% of weight measured
Agentic coding
40
57% measured
Long-horizon autonomy
49
30% measured
Knowledge-work agents
32
20% measured
Computer use
29
estimated
Security & context learning
44
50% measured

Price

Input
$1.14/M
Output
$9.09/M
Cache read
$0.11/M
3:1 blend
$3.13/M
Agent-effective
$0.69/M

Cheapest host for agents: OpenAI at $0.38/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verified68.0%8%source
Terminal-Bench 2.047.6%8%Terminus 2
DeepSWEest.6%predicted 38
FrontierSWEest.5%predicted 9
CursorBenchest.4%predicted 33
GSO-Bench13.7%4%OpenHands
METR Time Horizonsest.10%predicted 71
Vending-Bench 2$1,4736%source
PostTrainBenchest.4%predicted 22
APEX-Agentsest.8%predicted 36
GDPvalest.7%predicted 40
Remote Labor Indexest.5%predicted 2
DeepResearch Bench42.8%5%source
OSWorld 2.0est.6%predicted 6
OSWorldest.4%predicted 65
Cybenchest.5%predicted 64
CL-bench23.7%5%source

Other benchmarks

BenchmarkCategoryScoreSource
ARC-AGIReasoning72.8%source
ARC-AGI-2Reasoning17.6%source
Chess PuzzlesReasoning42.0%source
CritPtReasoning4.9%source
DTBenchKnowledge90.1%source
GPQA DiamondReasoning87.6%source
Humanity's Last ExamReasoning23.7%source
LMCAKnowledge43.9%source
OTIS Mock AIMEMath88.6%source
SciCodeCoding43.3%source
SimpleBenchReasoning53.2%source
SimpleQA VerifiedKnowledge48.0%source
VPCTMultimodal58.7%source
WebDev ArenaCoding1392source

API hosts

Live prices and uptime for every host serving GPT-5.1 via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
OpenAI$0.63$5.00$0.063$0.38100.0%—
Azure$1.25$10.00$0.13$0.76100.0%—
OpenAI$1.25$10.00$0.13$0.76100.0%—
OpenAI$2.50$20.00$0.25$1.51——