Qwen3.8 2.4T A95B
Neural Network
Qwen3.8 2.4T A95B API is an open Qwen language model with a sparse MoE architecture for text tasks.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 1 048 576 tokens
- Max answer length
- 131 072 tokens
- Reasoning effort
- low, medium, extra high
- Capabilities
- reasoning, function calling
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Qwen3.8 2.4T A95B?
Qwen3.8 2.4T A95B is an open-weights model from Qwen built on a sparse mixture of experts architecture: out of 2.4 trillion parameters, 95 billion are activated per request, combining large scale with efficient processing. It is the open version of Qwen3.8 Max from the same line. Via BotHub, the model is accessible from Russia without a VPN or foreign card; payment is in rubles per token used, and unused tokens do not expire. Over 250 models are available in one window, allowing you to compare responses and switch between them without changing interfaces. A unified OpenAI-compatible API connects Qwen to your service, allowing you to swap models without rewriting integrations. Dialogues are transmitted with AES-GCM encryption; for companies, we offer contracts, invoices, EDI, and an admin panel with limits. Use cases: parsing and summarizing long documents, drafting emails and work materials, answering standard user questions within your product, experimenting with an open model on your own data, and multi-model assemblies where Qwen handles part of the request flow.Frequently asked questions about Qwen3.8 2.4T A95B
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-based Qwen model with reasoning before answering. It accepts text, documents, and images as input, calls functions, and outputs structured JSON. It is suitable for working with code—refactoring, debugging, parsing large projects—as well as document analytics and STEM tasks where a chain of thought is important.
In a single response, the model outputs up to 131,072 tokens—enough for a large code module, a detailed technical report, or parsing an entire document. The context is 1,000,000 tokens, so you can upload source code or a package of files and get a long result without splitting it into parts.
Yes, reasoning before answering is confirmed: the model first builds a solution path and then provides the result. This helps with math, logic problems, and multi-step code debugging. Use it where result accuracy is more important than response speed—via our interface or API.
Yes, the model supports function calling and structured JSON output. You describe the tools, it selects the right one and returns arguments according to the specified schema—convenient for agents, document parsing, and integrations. The API in BotHub is OpenAI-compatible, so you can switch models without rewriting code.
Images and documents—yes: send a screenshot of an interface, a diagram, a table, a PDF, or a specification and ask for an analysis. There is no mention of audio or video in our data, so count on text, images, and files. The model returns the result as text.
We do not have data on prompt and response languages, so we will not make specific promises. It is fairer to test it on your own task: open the model on the website or in the Telegram bot, provide a typical request, and compare it with another model in the same window. Access from Russia without VPN, payment with Russian cards.