Qwen3.6 35B A3B
Neural Network
Qwen3.6 35B A3B API: Alibaba Cloud's multimodal open-weight model on a sparse MoE architecture with 3B active parameters.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 262 144 tokens
- Max answer length
- 235 929 tokens
- Reasoning effort
- low, medium, high
- Capabilities
- reasoning, function calling
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Qwen3.6 35B A3B?
Qwen3.6 35B A3B is a multimodal open-weight model from Alibaba Cloud. It is based on a hybrid sparse mixture-of-experts architecture: out of 35 billion parameters, only 3 billion are activated per token, distributing the load among experts rather than running through the entire network. With BotHub, you can run it without a VPN or foreign card: pay with Russian cards in rubles based on actual usage, and unused tokens do not expire. Other models are available in the same window, and a unified OpenAI-compatible API allows you to switch between them without rewriting integrations, with chats encrypted via AES-GCM. Teams have access to contracts, invoices, electronic document management, and an admin panel with employee limits. It is useful for analyzing texts and documents, drafting and editing, solving analytical tasks, integrating Qwen3.6 into your own services via API, and building multimodal scenarios where some requests are routed to this model.Frequently asked questions about Qwen3.6 35B A3B
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is a text model with a reasoning mode: code analysis and writing, refactoring, debugging, analytics, STEM tasks, and working with long documents. It accepts text, documents, and images as input and supports function calling. A 262,144-token context allows you to keep a large project or a collection of files in a single conversation.
The output limit is 235,929 tokens per response, which is a very large amount of text: technical documentation, detailed analysis, or an entire large code module. In practice, you set the length: ask it to shorten or continue — the 262,144-token context is enough to ensure nothing is lost.
Yes, reasoning before answering is supported: the model first works through the steps and then provides the result. This significantly helps with math, logic, code debugging, and tasks where multiple conditions need to be checked at once. It does not interfere with simple queries — the model answers directly.
Yes. The model supports function calling and structured JSON output, making it easy to integrate into agents, chatbots, and pipelines where the response goes directly into code. In BotHub, this is available via a unified OpenAI-compatible API: you can switch models without rewriting your integration.
Images and documents — yes, you can send them as input: the model will analyze a screenshot, diagram, table, or file and respond with text. Audio and video are not marked as supported inputs in our data. The output is text, and there are separate models in the catalog for audio and video.
We do not have specific benchmarks for Russian, so we will not provide ratings — it is more reliable to test it on your own task. Send a typical request and check the result; if you are not satisfied, switch to another model in the same window. Access from Russia without a VPN, payment in rubles.