Qwen3.8 2.4T A95B is a text-only model from Alibaba Cloud with supported function calling, structured outputs, web search, prefix completion, context caching, and batch inference in Model Studio. The provider catalogue also marks it as supporting reasoning and streaming.
It is described as a fine-grained mixture-of-experts model with hybrid attention and a very long context design, and Alibaba documents a maximum input length of 991,808 tokens and a maximum output length of 131,072 tokens.
| Developer | Alibaba Cloud |
|---|---|
| Model ID | alibaba/qwen3.8-2.4t-a95b |
| Release Date | August 11, 2026 |
| Context Window | 262,144 tokens (~262K) |
| Max Output Tokens | 52,429 tokens |
| Modalities | Text → Text |
| Capabilities | reasoning, streaming, structured (JSON schema) output, tool/function calling |
Model modality
The model supports **text input** and **text output** only in Alibaba Cloud’s Model Studio listing.
Reasoning behavior
The Hugging Face model page states that the model requires thinking mode for all interactions, and Alibaba’s catalogue lists it as supporting reasoning.
Tool and output support
Alibaba Cloud lists **function calling** and **structured outputs** as supported capabilities, and also lists web search support.
Long-context design
Alibaba documents a **1,000,000-token context length**, with a maximum input length of 991,808 tokens and a maximum output length of 131,072 tokens.
Architecture description
NVIDIA describes the model as a **fine-grained mixture-of-experts** system with a hybrid of full and linear attention, designed for demanding reasoning and agentic workloads.
Complex reasoning
The model is positioned for demanding reasoning workloads and is described as supporting reasoning in the provider catalogue.
Tool-using workflows
Alibaba Cloud lists function calling, structured outputs, and web search support, which fits agentic or tool-augmented applications.
Long-document processing
Its very large context limits make it suitable for processing long inputs and extended conversation state.
Agentic tasks
NVIDIA characterizes the model as designed for agentic workloads, indicating suitability for multi-step assistant and automation scenarios.
Qwen3.8-Max
Hugging Face discussion and third-party coverage describe Qwen3.8 2.4T A95B as the open build associated with the Qwen3.8 Max line, but those sources do not provide a direct side-by-side product comparison in the search results.
Sources: https://developer.nvidia.com/blog/serve-qwen3-8-2-4t-a95b-a-2-4t-parameter-model-with-configurable-reasoning-on-nvidia-gb300-nvl72/ · https://www.alibabacloud.com/help/en/model-studio/qwen3-8-2-4t-a95b · https://www.mindstudio.ai/blog/qwen3-8-2-4t-a95b-release
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Qwen3.8 2.4T A95B is documented as suited to coding, including autonomous coding and other hard coding tasks. It is also suited to large-scale document analysis and professional work. The model is documented for research and long-horizon agentic tasks, including long-running multi-step workflows and stronger autonomous planning.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "alibaba/qwen3.8-2.4t-a95b", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "alibaba/qwen3.8-2.4t-a95b", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"alibaba/qwen3.8-2.4t-a95b","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
This model’s documented pricing caveat is that Alibaba Cloud’s page shows the original API pricing only, excluding limited-time promotions, and it lists separate rates for cache hits and explicit cache creation[2
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Intelligence | 40 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Coding | 71.9 | Composite score across standardised coding evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Qwen3.8 2.4T A95B This page | Agentic workflows and structured output |
Qwen3.8 2.4T A95B has a 262,144 tokens context window and can return up to 52,429 tokens.
Qwen3.8 2.4T A95B takes text as input and returns text.
Use alibaba/qwen3.8-2.4t-a95b as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
Qwen3.8 2.4T A95B became available on August 12, 2026.
Qwen3.8 2.4T A95B is priced at input $3.4385 / 1M tokens, output $8.59625 / 1M tokens, cached input $0.6877 / 1M tokens.
Yes, Qwen3.8 2.4T A95B can stream responses as they are generated.
Qwen3.8 2.4T A95B was built by Alibaba Cloud.
It is a text-only model, accepting text input and producing text output.
It is built for reasoning-focused chat with tool use and structured output support, and reasoning is always enabled.
