SShogo Bench

Grok 4.5

xAI's Grok model for chat, coding, agentic tools, and lower hallucination risk

Shogo Agent Index

Estimated#36 of 92
49.4range 40.9–57.9 · 24% of weight measured
Agentic coding
54
17% measured
Long-horizon autonomy
59
50% measured
Knowledge-work agents
42
32% measured
Computer use
36
estimated
Security & context learning
48
estimated

Price

Input
$2.00/M
Output
$6.00/M
Cache read
$0.30/M
3:1 blend
$3.00/M
Agent-effective
$0.81/M

Cheapest host for agents: xAI at $0.81/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 78
Terminal-Bench 2.0est.8%predicted 72
DeepSWE53.8%6%mini-swe-agent · $2.42/task
FrontierSWEest.5%predicted 16
CursorBenchest.4%predicted 38
GSO-Benchest.4%predicted 29
METR Time Horizonsest.10%predicted 75
Vending-Bench 2$3,8876%source
PostTrainBench22.6%4%Cursor CLI
APEX-Agents56.2%8%source
GDPvalest.7%predicted 47
Remote Labor Indexest.5%predicted 4
DeepResearch Benchest.5%predicted 48
OSWorld 2.0est.6%predicted 10
OSWorldest.4%predicted 76
Cybenchest.5%predicted 75
CL-benchest.5%predicted 22

Other benchmarks

BenchmarkCategoryScoreSource
ARC-AGIReasoning87.2%source
ARC-AGI-2Reasoning52.6%source
Chess PuzzlesReasoning35.0%source
CritPtReasoning15.4%source
DTBenchKnowledge96.5%source
FrontierCodeCoding42.4%grok-build
FrontierMath Tiers 1-3Math57.2%source
FrontierMath Tier 4Math24.4%source
GPQA DiamondReasoning93.4%source
LMCAKnowledge45.2%source
MirrorCodeCoding32.2%source
OTIS Mock AIMEMath97.8%source
ProofBenchMath31.0%source
SciCodeCoding55.0%source
SimpleBenchReasoning70.0%source
SimpleQA VerifiedKnowledge48.3%source
WebDev ArenaCoding1555source

API hosts

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

HostInputOutputCache readAgent $/MUptime 24hQuant
xAI$2.00$6.00$0.30$0.81100.0%—
xAI$2.00$6.00$0.30$0.8199.9%—
xAI$4.00$12.00$0.60$1.63——
xAI$4.00$12.00$0.60$1.63——