SShogo Bench

Qwen3 32B

Qwen instruction model for multilingual chat, reasoning, and tool use

Shogo Agent Index

Unranked

Not ranked yet: 0 agentic benchmarks measured, and we need at least 2 plus an overall capability score. It will be ranked as results are published.

Agentic coding
19
estimated
Long-horizon autonomy
30
estimated
Knowledge-work agents
18
estimated
Computer use
8
estimated
Security & context learning
19
estimated

Price

Input
$0.18/M
Output
$0.70/M
Cache read
—/M
3:1 blend
$0.31/M
Agent-effective
$0.20/M

Cheapest host for agents: DeepInfra at $0.070/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 53
Terminal-Bench 2.0est.8%predicted 11
DeepSWEest.6%predicted 11
FrontierSWEest.5%predicted 1
CursorBenchest.4%predicted 19
GSO-Benchest.4%predicted 2
METR Time Horizonsest.10%predicted 53
Vending-Bench 2est.6%predicted 3
PostTrainBenchest.4%predicted 12
APEX-Agentsest.8%predicted 13
GDPvalest.7%predicted 19
Remote Labor Indexest.5%predicted 0
DeepResearch Benchest.5%predicted 42
OSWorld 2.0est.6%predicted 1
OSWorldest.4%predicted 19
Cybenchest.5%predicted 24
CL-benchest.5%predicted 13

Other benchmarks

BenchmarkCategoryScoreSource
Aider PolyglotCoding40.0%source
CritPtReasoning0.3%source
DTBenchKnowledge67.5%source
Fiction.LiveBenchLong context74.2%source
GPQA DiamondReasoning65.7%source
LMCAKnowledge17.3%source
OTIS Mock AIMEMath66.9%source
SciCodeCoding35.4%source

API hosts

Live prices and uptime for every host serving Qwen3 32B via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
DeepInfra$0.064$0.22—$0.07096.0%fp8
DeepInfra$0.080$0.28—$0.09098.8%fp8
Tenstorrent$0.14$0.40$0.090$0.1197.2%fp8
DeepInfra$0.12$0.42—$0.1399.9%fp8
SiliconFlow$0.14$0.57—$0.1692.4%fp8