SShogo Bench

Inkling

Multimodal MoE reasoning model (975B total, 41B active) for text, image, and audio

Shogo Agent Index

Estimated#64 of 92
35.8range 23–48.6 · 13% of weight measured
Agentic coding
38
14% measured
Long-horizon autonomy
44
estimated
Knowledge-work agents
30
32% measured
Computer use
25
estimated
Security & context learning
38
estimated

Price

Input
$1.87/M
Output
$4.68/M
Cache read
$0.37/M
3:1 blend
$2.57/M
Agent-effective
$0.79/M

Cheapest host for agents: Together at $0.47/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 70
Terminal-Bench 2.0est.8%predicted 47
DeepSWEest.6%predicted 32
FrontierSWE4.1%5%proximus · $9.15/task
CursorBenchest.4%predicted 30
GSO-Benchest.4%predicted 13
METR Time Horizonsest.10%predicted 68
Vending-Bench 2est.6%predicted 21
PostTrainBenchest.4%predicted 20
APEX-Agents33.8%8%source
GDPvalest.7%predicted 35
Remote Labor Indexest.5%predicted 2
DeepResearch Benchest.5%predicted 46
OSWorld 2.0est.6%predicted 4
OSWorldest.4%predicted 56
Cybenchest.5%predicted 57
CL-benchest.5%predicted 18

Other benchmarks

BenchmarkCategoryScoreSource
ARC-AGIReasoning79.5%source
ARC-AGI-2Reasoning36.5%source
Chess PuzzlesReasoning21.0%source
CritPtReasoning5.4%source
DTBenchKnowledge87.5%source
FrontierCodeCoding14.0%mini-swe-agent
FrontierMath Tiers 1-3Math33.3%source
FrontierMath Tier 4Math4.9%source
GPQA DiamondReasoning88.3%source
LMCAKnowledge37.6%source
OTIS Mock AIMEMath88.9%source
ProofBenchMath0.0%source
SciCodeCoding47.0%source
SimpleBenchReasoning50.0%source
SimpleQA VerifiedKnowledge40.3%source
WebDev ArenaCoding1410source
WeirdML v3Coding3.9%opencode

API hosts

Live prices and uptime for every host serving Inkling via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
Together$1.00$4.05$0.17$0.47100.0%—