SShogo Bench

Muse Spark Open

Muse Spark Open is an 816B total / 42B active open-weight multimodal reasoning model from Meta for coding and long-horizon agentic work, with 1M-token context.

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
54
estimated
Long-horizon autonomy
58
estimated
Knowledge-work agents
35
32% measured
Computer use
37
estimated
Security & context learning
49
estimated

Price

Input
$1.40/M
Output
$4.65/M
Cache read
$0.17/M
3:1 blend
$2.21/M
Agent-effective
$0.56/M

Cheapest host for agents: BaseTen at $0.50/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 74
DeepSWEest.6%predicted 50
FrontierSWEest.5%predicted 17
CursorBenchest.4%predicted 38
GSO-Benchest.4%predicted 31
METR Time Horizonsest.10%predicted 76
Vending-Bench 2est.6%predicted 51
PostTrainBenchest.4%predicted 26
APEX-Agents33.4%8%source
GDPvalest.7%predicted 48
Remote Labor Indexest.5%predicted 4
DeepResearch Benchest.5%predicted 48
OSWorld 2.0est.6%predicted 10
OSWorldest.4%predicted 78
Cybenchest.5%predicted 76
CL-benchest.5%predicted 22

Other benchmarks

BenchmarkCategoryScoreSource
ARC-AGIReasoning89.0%source
ARC-AGI-2Reasoning57.5%source
CritPtReasoning14.9%source
FrontierMath Tiers 1-3Math69.8%source
FrontierMath Tier 4Math31.7%source
GPQA DiamondReasoning88.6%source
OTIS Mock AIMEMath98.1%source
SciCodeCoding58.4%source
SimpleQA VerifiedKnowledge56.6%source

API hosts

Live prices and uptime for every host serving Muse Spark Open via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
BaseTen$1.25$4.25$0.15$0.50100.0%fp4
Fireworks$1.25$4.25$0.15$0.50100.0%—
Modal$1.25$4.25$0.15$0.50100.0%—
Nebius$1.25$4.25$0.15$0.5097.9%fp4
Together$1.25$4.25$0.15$0.5099.8%—