All AI models · LLM benchmarks · Methodology

Alibaba

Qwen3.5 397B-A17B FP8

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

Model facts

Context window
2K tokens
Maximum output
66K tokens
Cheapest paid input
$0.99 per 1M tokens
Cheapest paid output
$4.46 per 1M tokens
Open weights
yes
Providers
1

Capabilities and modalities

reasoning, tool calling, structured output, attachments, temperature control, open weights, text, image.

Local hardware estimate

At 4K context: Q4_K_M 214.5 GiB, Q8_0 402.4 GiB, F16 754.6 GiB working memory.

Parameters
403.4 billion
Architecture
qwen3_5moe
Model license
apache-2.0

Planning estimate adapted from the MIT-licensed whichllm estimator using metadata from Hugging Face; not a vendor minimum requirement.

Observed price history

  • 2026-08-24: $0.69 input / $4.14 output per 1M tokens
  • 2026-08-28: $0.99 input / $4.46 output per 1M tokens

API providers

  • Infomaniak — model id Qwen/Qwen3.5-397B-A17B-FP8; $0.99 input / $4.46 output per 1M tokens; provider documentation