SShogo Bench

DeepSeek V3.1

A large hybrid model that supports both thinking and non-thinking modes via prompt templates.

Shogo Agent Index

Unranked

Not ranked yet: 0 agentic benchmarks measured, and we need at least 2 plus an overall capability score. It will be ranked as results are published.

Agentic coding
24
estimated
Long-horizon autonomy
34
estimated
Knowledge-work agents
21
estimated
Computer use
13
estimated
Security & context learning
24
estimated

Price

Input
$0.50/M
Output
$1.50/M
Cache read
$0.20/M
3:1 blend
$0.75/M
Agent-effective
$0.30/M

Cheapest host for agents: DeepInfra at $0.19/M.

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

Agentic benchmarks

BenchmarkScoreNormalisedIndex weightHarness / source
SWE-bench Verifiedest.8%predicted 59
Terminal-Bench 2.0est.8%predicted 20
DeepSWEest.6%predicted 16
FrontierSWEest.5%predicted 2
CursorBenchest.4%predicted 22
GSO-Benchest.4%predicted 4
METR Time Horizonsest.10%predicted 58
Vending-Bench 2est.6%predicted 6
PostTrainBenchest.4%predicted 14
APEX-Agentsest.8%predicted 17
GDPvalest.7%predicted 24
Remote Labor Indexest.5%predicted 1
DeepResearch Benchest.5%predicted 43
OSWorld 2.0est.6%predicted 1
OSWorldest.4%predicted 30
Cybenchest.5%predicted 34
CL-benchest.5%predicted 15

Other benchmarks

BenchmarkCategoryScoreSource
DTBenchKnowledge82.7%source
Fiction.LiveBenchLong context52.8%source
LMCAKnowledge24.3%source
SimpleBenchReasoning40.0%source

API hosts

Live prices and uptime for every host serving DeepSeek V3.1 via OpenRouter.

HostInputOutputCache readAgent $/MUptime 24hQuant
DeepInfra$0.25$0.95$0.13$0.19100.0%fp4
CoreWeave$0.50$1.50$0.20$0.30100.0%fp8
SiliconFlow$0.27$1.00—$0.3099.7%fp8