Qwen 3.5 397B

CodeVisionReasoningWeb searchFunction calling
Advertised as qwen-3.5-397bqwen/qwen3.5-397b-a17bqwen3-5-397b-a17bqwen3.5-397bqwen3.5-397b-a17bQwen3.5-397B-A17B
Quick reference
Qwen 3.5 397B — TLDR
  • 🧩 397B MoE, only 17B active per token
  • 🧠 Flagship reasoning, coding, and general knowledge
  • 📏 128K-token context window
  • 👁️ Vision, function calling, and web search
  • 📜 Open-weight under Apache 2.0
  • 🎯 Alibaba's Qwen 3.5 generation flagship
💰 Best price on AntSeed
$0.151 / $0.90972%
per 1M · cheapest in / out
📏 Context
128K tokens
🐜 Sellers
12
advertising on AntSeed
Provider

Alibaba Group is a Chinese multinational technology company founded in 1999 and headquartered in Hangzhou, Zhejiang. Originally built around e-commerce and cloud computing, Alibaba has become one of the most prolific contributors to open-weight AI research, developing the Qwen…

Explore 39 more models by Alibaba Group
About this model

Qwen 3.5 397B is Alibaba's flagship reasoning model from the Qwen 3.5 generation, built on a Mixture-of-Experts architecture that totals 397 billion parameters while activating just 17 billion per token. Released in February 2026, it pairs strong chain-of-thought reasoning with broad capability coverage: coding, general knowledge, vision input, native function calling, and web search all sit within a 128,000-token context window. The sparse MoE design is the key trade-off here — it delivers frontier-scale reasoning quality while keeping inference cost closer to a mid-size dense model.

Within Alibaba's lineup it anchors the top of the Qwen 3.5 family, sitting above the lighter Qwen 3.5 35B A3B and Qwen 3.5 9B variants. The line has since continued with Qwen 3.6 27B (released April 2026), so this model represents the heavyweight peak of the prior generation rather than the newest small release. It ships under a permissive Apache 2.0 license, with strong adoption reflected in its Hugging Face download and like counts.

It is best suited for demanding agentic and analytical work — multi-step reasoning, complex code generation, tool-driven workflows, and long-document tasks — where its large expert pool and generous context justify reaching for a flagship over a smaller sibling.

View source on GitHub ↗View model card on HuggingFace ↗

This About section is AI-generated from public sources via VeniceStats + Venice inference, with no human editing. It may contain inaccuracies.

Usage on AntSeed
Tokens served
29.93M
input + output
Requests
442
settled calls
Buyers
9
distinct, on this model
Sellers used
6
of 12 advertising
Settled
$14.97
gross USDC, this model
Sellers serving Qwen 3.5 397B (12)compare on the network explorer →
SellerReputationRoutingInput $/MCached $/MOutput $/MCategoriesAPI

"Best price" and the seller table are live AntSeed catalog data (advertised $/1M tokens — or $ per generated image for unit-billed image models — not settled amounts). Reputation = on-chain trust (0-100). "Routing" = the SDK's default buyer routing (what the VPR desktop app ships with): a trust ≥ 60 gate on the effective reputation, then cheapest-first among routable sellers; live failover state (per-peer cooldowns) is buyer-side runtime and not included. Model knowledge (TLDR, provider, About) via the VeniceStats enrichment layer. Advertised catalog, not the model used in any specific purchase. "Usage on AntSeed" counts only settlements whose buyers share the per-model split on-chain (metadata v2/v3, opt-in), so every usage figure is a lower bound.