Qwen 3.8 Flash Next spotted, runs on solo or dual DGX Spark setups
lhoestq · x · 2026-09-21
MiaAIlab shared a preview of "Qwen 3.8 Flash Next" with deployment instructions for both a solo DGX Spark and dual DGX Sparks. Hugging Face co-founder Quentin Lhoest retweeted saying he's upgrading to try it. The model name suggests an unannounced Qwen release — treat as unconfirmed.
More from Models
- StepFun urgently scales up after Step 5 Preview surge, plans free Pro for PhD students — teortaxesTex · 2026-09-21
- M5 Ultra Mac Studio first impressions: 2x faster Rust builds than M3 Ultra — BenBajarin · 2026-09-21
- Is Everyone Adopting DeepSeek's Architectural Innovations? Redditors Weigh In — EveningNo8643 · 2026-09-21
- Weaviate Podcast #145: Humans& Researchers on Persimmon, a User Model That Predicts Human Turn-Taking — CShorten30 · 2026-09-21
- StepFun's Step 5 Preview knows when to stop, edging GLM 5.3 on long-horizon agent tasks — omarsar0 · 2026-09-21
- Jev, the first decision-only AI model, makes calls in 200ms for a fraction of a cent — Cole Medin · 2026-09-21