SShogo Bench

GPT-5.3 Codex

Coding-optimized GPT model for repository edits, reviews, and agentic software work

Shogo Agent Index

Supported#37 of 92
48.4range 38.5–58.4 · 32% of weight measured
Agentic coding
53
46% measured
Long-horizon autonomy
62
80% measured
Knowledge-work agents
37
estimated
Computer use
35
estimated
Security & context learning
47
estimated

Price

Input
$1.75/M
Output
$14.00/M
Cache read
$0.17/M
3:1 blend
$4.81/M
Agent-effective
$1.06/M

Cheapest host for agents: Azure at $1.06/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verified74.8%8%source
Terminal-Bench 2.078.4%8%SageAgent
DeepSWEest.6%predicted 47
FrontierSWEest.5%predicted 15
CursorBenchest.4%predicted 37
GSO-Benchest.4%predicted 27
METR Time Horizons74.5%10%source
Vending-Bench 2$5,9406%source
PostTrainBenchest.4%predicted 25
APEX-Agentsest.8%predicted 44
GDPvalest.7%predicted 46
Remote Labor Indexest.5%predicted 4
DeepResearch Benchest.5%predicted 48
OSWorld 2.0est.6%predicted 9
OSWorldest.4%predicted 75
Cybenchest.5%predicted 73
CL-benchest.5%predicted 21

Other benchmarks

BenchmarkCategoryScoreSource
CritPtReasoning16.9%source
WebDev ArenaCoding1370source

API hosts

Live prices and uptime for every host serving GPT-5.3 Codex via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
Azure$1.75$14.00$0.17$1.06100.0%—
OpenAI$1.75$14.00$0.17$1.06100.0%—
OpenAI$3.50$28.00$0.35$2.12——