Qwen3.8-Omni-Flash is the speed- and cost-optimised omni tier of Alibaba Cloud's Qwen3.8 family. It is natively multimodal on the way in — text, images, audio and video all go into the same prompt — and returns text. With a 1,000,000-token context window and hybrid thinking, it is aimed at high-volume multimodal understanding rather than generation.
Four input modalities, one prompt
Audio, video, images and text are handled natively, so a recording, a screenshot and the surrounding instructions can be sent together without a separate transcription or captioning step.
A million tokens of context
Long enough for whole repositories, large document sets or extended transcripts, with up to 131,072 tokens back.
Hybrid thinking, switchable per request
Reasoning is off by default and turned on with enable_thinking, with thinking_budget to cap how much of it you pay for. Reasoning tokens are billed at the output rate.
Built for cost
At $0.195 per million input tokens, with cached input an order of magnitude cheaper again, it is priced for workloads that run constantly rather than occasionally.
Reasoning tokens are charged at the output rate.
Multimodal understanding at volume
Classifying, summarising or extracting from mixed media where a frontier model would be overkill on price.
Audio and video analysis
Meeting recordings, calls and clips go in directly and come back as structured text.
Long-document work
The 1M window removes most of the need for chunking and retrieval scaffolding.
Agentic workflows
Parallel tool calling and structured output support agent loops that plan, call tools and iterate.
Source: https://api.aimlapi.com/v1/models
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "alibaba/qwen3.8-omni-flash", "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-omni-flash", "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-omni-flash","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
Qwen3.8 Omni Flash has a 1,000,000 token context window and can return up to 131,072 tokens.
It accepts audio, images, text and video as input and returns text. All four input types can go into the same prompt, so a recording or a clip does not need a separate transcription step first.
Qwen3.8 Omni Flash is priced at input $0.195 / 1M tokens, output $0.611 / 1M tokens, cached input $0.0208 / 1M tokens. Reasoning tokens are charged at the output rate.
No. Audio and video are input modalities only — the model listens and watches, but answers in text. There is no parameter on the chat completions endpoint to request an audio response.
Reasoning is off by default. Set enable_thinking to true to switch it on, and use thinking_budget to cap how many reasoning tokens a request may spend. Those tokens are billed at the output rate.
Yes. It supports function calling including parallel tool calls, structured (JSON schema) output, streaming and context caching.
Qwen3.8 Omni Flash was built by Alibaba Cloud as the speed- and cost-optimised omni tier of the Qwen3.8 family. It became available on September 17, 2026.
It is served on the OpenAI-compatible chat completions endpoint. Point any OpenAI SDK at https://api.aimlapi.com/v1 and set the model to alibaba/qwen3.8-omni-flash. The short alias qwen3.8-omni-flash also resolves to it.
