SShogo Bench

Qwen3 235B A22B Thinking 2507

Qwen reasoning model for deliberate problem solving, math, and coding

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
28
estimated
Long-horizon autonomy
34
30% measured
Knowledge-work agents
23
estimated
Computer use
16
estimated
Security & context learning
29
estimated

Price

Input
$0.20/M
Output
$0.60/M
Cache read
—/M
3:1 blend
$0.30/M
Agent-effective
$0.22/M

Cheapest host for agents: Venice at $0.60/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 63
Terminal-Bench 2.0est.8%predicted 27
DeepSWEest.6%predicted 21
FrontierSWEest.5%predicted 3
CursorBenchest.4%predicted 24
GSO-Benchest.4%predicted 6
METR Time Horizonsest.10%predicted 61
Vending-Bench 2$-116%source
PostTrainBenchest.4%predicted 16
APEX-Agentsest.8%predicted 21
GDPvalest.7%predicted 27
Remote Labor Indexest.5%predicted 1
DeepResearch Benchest.5%predicted 44
OSWorld 2.0est.6%predicted 2
OSWorldest.4%predicted 38
Cybenchest.5%predicted 41
CL-benchest.5%predicted 16

Other benchmarks

BenchmarkCategoryScoreSource
CritPtReasoning0.0%source
DTBenchKnowledge80.3%source
Fiction.LiveBenchLong context75.0%source
GPQA DiamondReasoning80.1%source
LMCAKnowledge29.3%source
OTIS Mock AIMEMath86.7%source
SciCodeCoding41.4%source
SimpleQA VerifiedKnowledge40.4%source

API hosts

Live prices and uptime for every host serving Qwen3 235B A22B Thinking 2507 via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
Venice$0.45$3.50—$0.6099.9%fp8