SShogo Bench

Gemini 4 Hydrogen

Google's Gemini 4 model for everyday coding and agentic workflows, with four thinking levels and a 1M-token context window

Shogo Agent Index

Estimated#7 of 92
69.4range 61.8–77.1 · 23% of weight measured
Agentic coding
78
26% measured
Long-horizon autonomy
78
30% measured
Knowledge-work agents
58
32% measured
Computer use
61
estimated
Security & context learning
61
estimated

Price

Input
$1.50/M
Output
$7.50/M
Cache read
$0.075/M
3:1 blend
$3.00/M
Agent-effective
$0.63/M

Cheapest host for agents: Google AI Studio at $0.32/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 88
Terminal-Bench 2.0est.8%predicted 95
DeepSWEest.6%predicted 78
FrontierSWE53.1%5%proximus · $61.79/task
CursorBench57.8%4%source · $5.86/task
GSO-Benchest.4%predicted 72
METR Time Horizonsest.10%predicted 86
Vending-Bench 2$14,5496%source
PostTrainBenchest.4%predicted 38
APEX-Agents78.0%8%source
GDPvalest.7%predicted 68
Remote Labor Indexest.5%predicted 17
DeepResearch Benchest.5%predicted 52
OSWorld 2.0est.6%predicted 39
OSWorldest.4%predicted 95
Cybenchest.5%predicted 93
CL-benchest.5%predicted 29

Other benchmarks

BenchmarkCategoryScoreSource
Chess PuzzlesReasoning60.0%source
FrontierMath Tiers 1-3Math88.1%source
FrontierMath Tier 4Math87.8%source
GPQA DiamondReasoning96.8%source
OTIS Mock AIMEMath100.0%source
SimpleQA VerifiedKnowledge70.9%source

API hosts

Live prices and uptime for every host serving Gemini 4 Hydrogen via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
Google AI Studio$0.75$3.75$0.037$0.3298.6%—
Google$1.50$7.50$0.075$0.6390.5%—
Google AI Studio$1.50$7.50$0.075$0.6397.0%—
Google$2.70$13.50$0.14$1.14100.0%—
Google AI Studio$2.70$13.50$0.14$1.14100.0%—