Qwen3.8 Omni Flash

Alibaba's Qwen team has released Qwen3.8-Omni-Flash, its first omni-modal model built around agentic capabilities. It accepts text, images, audio, and video inputs and returns text only, unifying audio-video understanding, reasoning, and tool use in one model. The architecture builds on the Qwen3.8-Flash-Next base that shipped with open weights in August 2026. The model offers a 1M-token context window, with QwenCloud listing 991K max input and 131K max output, plus 262K max reasoning length. Thinking is on by default with reasoning effort set to xhigh. It is available as a hosted API on QwenCloud, Alibaba Cloud Model Studio, and Qwen Studio, though no open weights were announced at launch.

Alibaba's Qwen team released Qwen3.8-Omni-Flash as the next-generation omnimodal model replacing Qwen3.5-Omni-Plus, sitting in the Flash tier as a cost-efficient, high-throughput option. It processes text, image, audio, and video in one model with a 1M-token context window, aiming for agent capability in audio and video rather than just understanding. The model claims audio-visual performance close to Gemini 3.8 Flash and overall audio performance exceeding it, with a reported 25%+ average improvement over its predecessor across 29 evaluations. Headline agent gains include a 36.5-point lead on WildClawBench-MM and a reported 98%+ drop in audio input per-hour price, positioned for audio- or video-heavy workloads.

Qwen3.8-Omni-Flash was benchmarked across audio-visual agent and understanding suites with exact tabulated figures from the official launch post. It scored 71.0 on WildClawBench-MM (multimodal tool use with Claude Code harness), 69.6 on UniClawBench (OpenClaw harness), 36.8 on AgenticVBench, and 74.0 on OmniGAIA, trailing Gemini 3.8 Flash on the latter two. On audio-visual reasoning benchmarks, the model scored 63.4 on OmniVideoBench and 65.0 on Video-MME-v2. Using the Qwen Code agent harness on OmniVideoBench raised the score to 67.8 while reducing tokens per query from 145,736 to 79,117, indicating significant efficiency gains from agentic integration.