llm.ing

Qwen3.5 397B-A17B

Alibabaopen weightsOpenRouter ↗Hugging Face ↗Artificial Analysis ↗

Large open Qwen multimodal MoE for visual agents and long technical tasks

Benchmarks

BenchmarkCategoryScoreVariantProvenanceSourceObserved
AA Agentic Indexagentic8.3—mirroropenrouterSep 20, 2026
8.3—independentaaSep 20, 2026
AA Coding Indexcoding48.2—mirroropenrouterJul 23, 2026
48.2—independentaaJul 24, 2026
AA Intelligence Indexintelligence21.4non-reasoningindependentaaSep 9, 2026
18.4—independentaaSep 20, 2026
21.4—mirroropenrouterOct 5, 2026
Design Arena — 3dpreference1134—mirroropenrouterNov 19, 2026
Design Arena — codecategoriespreference1164—mirroropenrouterNov 19, 2026
Design Arena — datavizpreference1170—mirroropenrouterNov 18, 2026
Design Arena — gamedevpreference1123—mirroropenrouterNov 19, 2026
Design Arena — svgpreference1125—mirroropenrouterNov 19, 2026
Design Arena — uicomponentpreference1150—mirroropenrouterNov 19, 2026
Design Arena — websitepreference1180—mirroropenrouterNov 19, 2026
LMArena Textpreference1438—crowdlmarenaNov 9, 2026
LMArena Visionpreference1262—crowdlmarenaNov 9, 2026
LMArena WebDevpreference1400—crowdlmarenaNov 11, 2026
Median output speedspeed58 tok/s—independentaaNov 19, 2026
58 tok/snon-reasoningindependentaaNov 19, 2026
Median time to first tokenspeed2.10s—independentaaNov 18, 2026
2.20snon-reasoningindependentaaNov 19, 2026

Includes index data from Artificial Analysis.

Providers

Live operational stats per hosting endpoint, via OpenRouter.

ProviderQuantContext$/1M in · outUptime 24hLatencyThroughput
Alibaba—262K$0.39 · $2.3498.40%——
AtlasCloudfp8262K$0.55 · $3.596.99%——
DeepInfrafp8262K$0.45 · $3100.00%——
DeepInfrafp8262K$0.36 · $2.4100.00%——
DeepInfrafp8262K$0.675 · $4.597.34%——
GMICloudfp8262K$0.6 · $3.659.52%——
Novita—262K$0.6 · $3.694.14%——
Parasailfp8262K$0.5 · $3.6100.00%——
StreamLake—256K$0.6 · $3.696.45%——
Venice—128K$0.75 · $4.596.46%——