Qwen3.8-Flash-Next shows strong agentic results

Qwen3.8-Flash-Next scores 61.1 out of 100 and ranks #67 of 230 on BenchLM’s public leaderboard, with 24 source-displayable benchmark rows. The model’s strongest eligible category is Agentic at #8, where the profile says it is particularly useful for coding agents, browser research, and computer-use workflows.

Category coverage remains uneven. Qwen3.8-Flash-Next ranks #26 of 148 in coding and #10 of 43 in instruction following, while reasoning, math, and multilingual categories are not measured. Multimodal and knowledge rows have verified source-displayable evidence, but the public category table does not assign ranks there.

Alibaba released the model on Aug 26, 2026 as an open-weight checkpoint under the Qwen Community 1.0 license. The profile lists a 262K context window, native text, image, and video input in the model card, and a 262,144-token native context window that can be extended to 1,000,000 tokens with YaRN. Runtime is listed at 86tok/s with First token 25.86 s, while hosted token pricing is not available because deployment is self-hosted and infrastructure cost varies.