SShogo Bench

Claude Haiku 4.5

Fast Claude model for responsive assistance, classification, and lightweight agents

Shogo Agent Index

Estimated#76 of 92
27.1range 18.9–35.3 · 19% of weight measured
Agentic coding
30
23% measured
Long-horizon autonomy
34
30% measured
Knowledge-work agents
23
20% measured
Computer use
15
estimated
Security & context learning
28
estimated

Price

Input
$1.00/M
Output
$5.00/M
Cache read
$0.10/M
3:1 blend
$2.00/M
Agent-effective
$0.46/M

Cheapest host for agents: Amazon Bedrock at $0.46/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 62
Terminal-Bench 2.035.5%8%Goose
DeepSWEest.6%predicted 20
FrontierSWEest.5%predicted 3
CursorBenchest.4%predicted 24
GSO-Benchest.4%predicted 6
METR Time Horizonsest.10%predicted 61
Vending-Bench 2$4596%source
PostTrainBenchest.4%predicted 15
APEX-Agentsest.8%predicted 20
GDPvalest.7%predicted 26
Remote Labor Indexest.5%predicted 1
DeepResearch Bench45.5%5%source
OSWorld 2.0est.6%predicted 2
OSWorldest.4%predicted 36
Cybenchest.5%predicted 40
CL-benchest.5%predicted 16

Other benchmarks

BenchmarkCategoryScoreSource
ARC-AGIReasoning47.7%source
ARC-AGI-2Reasoning4.0%source
BALROGLong-horizon31.2%source
Chess PuzzlesReasoning8.0%source
CritPtReasoning0.0%source
DTBenchKnowledge73.6%source
ExploitBenchSecurity & context13.7%source
GPQA DiamondReasoning71.2%source
LMCAKnowledge30.9%source
MATH Level 5Math96.4%source
OTIS Mock AIMEMath66.7%source
SciCodeCoding42.2%source
SimpleQA VerifiedKnowledge13.2%source
WebDev ArenaCoding1329source

API hosts

Live prices and uptime for every host serving Claude Haiku 4.5 via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
Amazon Bedrock$1.00$5.00$0.10$0.46100.0%—
Anthropic$1.00$5.00$0.10$0.46100.0%—
Azure$1.00$5.00$0.10$0.46100.0%—
Google$1.00$5.00$0.10$0.46100.0%—