Alibaba (Qwen)

Qwen3.8-Omni-Flash-Realtime

Released Sep 18, 2026

The realtime variant of Alibaba's Qwen3.8-Omni-Flash omnimodal model. It perceives and responds while receiving live audio and video streams, and can call tools along the way.

Price per minute: —
Pricing: Singapore, per 1M tokens: input text/image/video $0.23, audio $0.93; output text $0.70, audio $1.87.
Inputs: audio, text, image, video
Tool calling: Yes
Languages: 74
Voices: —
Big Bench Audio: —
Access: Generally available
Licence: Proprietary

Strengths & weaknesses

Strengths

Live audio and video input, plus images and text
Speech recognition in 74 languages and 39 Chinese dialects
Custom function calling, remote MCP and web search
Audio at $0.93 in / $1.87 out per 1M tokens

Weaknesses

Speech output in fewer languages (29 plus 7 dialects) than it understands
Not yet on the AA leaderboard

Not notable: not on AA. releaseDate is the qwen.ai blog date (2026/09/18, seen via search); the qwen.ai page is JS-rendered and didn't load, and the Alibaba Cloud Community mirror is dated 2026-09-20. Language counts conflict: the blog says 74 languages + 39 dialects recognised and 29 + 7 generated, while the Model Studio realtime doc says 113 languages/dialects recognised and 36 generated. languages = 74 (blog recognition). Open weights not stated.

Links

API model strings

alibabaqwen3.8-omni-flash-realtime