Aleph Alpha's open-weight English-German mixture-of-experts model for self-hosting, released under Apache 2.0 in October 2026
Qwen3.8-Omni-Flash-Realtime
New· 4Alibaba's streaming audio and video conversation model, with spoken replies, tool calling and remote MCP through its API
About Qwen3.8-Omni-Flash-Realtime
When an assistant has to listen and answer aloud, the delay between a person finishing a sentence and hearing a reply decides whether it feels like a conversation. Qwen3.8-Omni-Flash-Realtime is the streaming edition that Alibaba derived from its Qwen3.8-Omni-Flash omnimodal model, tuned for that kind of low-latency exchange.
Two dates for one release
The Qwen team announced the model on 18 September 2026, alongside Qwen3.8-Omni-Flash, and included API sample code. Alibaba Cloud Model Studio's lifecycle page gives 21 September 2026 as the release date, for the Singapore and China (Beijing) regions.
Streaming in, speech out
The model takes streaming audio and images, including video frames at a recommended rate of one frame per second, and replies in text and speech. Clients can connect over WebSocket, WebRTC or Alibaba's own AOQ protocol. Among the listed capabilities are custom function calling, remote MCP tools (which need approval unless configured otherwise), audio input on two or four channels, video aggregation to cut computation, voice cloning, and spatial-audio perception that Alibaba says can locate where a sound comes from. Recognition covers 113 languages and dialects; generated speech covers 36. Alibaba's own tests put the time to the first audio packet at around one second for short audio inputs.
Limits to design around
Total input is capped at 196,608 tokens and a WebSocket session lasts at most 120 minutes. The running context holds at most 100 audio turns or ten minutes of audio. Web search and tool calling are mutually exclusive. Each second of input audio counts as 7 tokens, and each second of output audio as 12.5, and spoken replies are billed for the audio and for its transcript. A free quota applies in Singapore only.
Access
The model itself is proprietary and reachable only through the API. What Alibaba has open-sourced is a companion runtime, Qwen-Live Harness, not the weights. Uses it suggests are live customer service, speaking practice and assistants driven by a camera.
TrustList summary
Qwen3.8-Omni-Flash-Realtime is Alibaba's streaming variant of the Qwen3.8-Omni-Flash omnimodal model, built for conversation with low delay. The Qwen team announced it on 18 September 2026, and Alibaba Cloud Model Studio lists its release as 21 September 2026 in Singapore and China (Beijing), so the sources differ on the date.
What Qwen3.8-Omni-Flash-Realtime does
- Streaming input: takes streamed audio and images, with video sent as frames and one per second recommended.
- Spoken replies: answers with text and speech, with speech recognition in 113 languages and dialects and speech generation in 36.
- Tools: custom function calling and remote MCP tools, with approval required by default for MCP calls.
- Connections: WebSocket, WebRTC or Alibaba's AOQ protocol, with voice cloning and spatial-audio perception among its listed features.
Customers and markets
Alibaba suggests live customer service, speaking practice and camera-based assistants.
Key facts
- Availability: API only. The model is proprietary, but the companion runtime Qwen-Live Harness is open source.
- Input limit: 196,608 tokens in total, with WebSocket sessions lasting at most 120 minutes.
- Context: up to 100 audio turns or 600 seconds of audio.
Summary written by TrustList from the business's own published information, Nov 2026.
Benchmarks & AI stats
| Benchmark | Official | Community avg |
|---|---|---|
| No benchmark scores yet. Be the first to add one. | ||
“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).
Sign in to add a benchmark score for this model.
Qwen3.8-Omni-Flash-Realtime launches
Trust Score
An earned signal from verification, reviews, awards, transparency and engagement - the vendor can't buy it.
Updated 9/22/2026
Request a demo or quote from Qwen3.8-Omni-Flash-Realtime
Reviews
Write the first review of Qwen3.8-Omni-Flash-Realtime
Used it? Your experience helps other buyers decide.
Write a reviewQuestions & answers
No questions yet. Be the first to ask about Qwen3.8-Omni-Flash-Realtime.
Qwen3.8-Omni-Flash-Realtime alternatives
Similar listings buyers compare against this one.
Cantina Security's first open-weight security model, a post-trained GLM-5.3-Flash for vulnerability research and exploit checks
Microsoft AI's text-to-speech model for keeping one voice across 23 languages, with voice cloning from short audio
JetBrains' open 12B mixture-of-experts coding model with 2.5B parameters active per token under Apache 2.0
Open-weight translation model from Bilibili's Index team, covering 150 languages and following glossary and style instructions
Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure
Liquid AI's open-weight decision model, giving yes or no, choice or score answers from text and images
Anthropic's smallest and fastest Claude 5.5 model, available on its own API and three cloud platforms
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- Write a reviewHelp other buyers decide
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.