Qwen 3.6 weights missing, LocalLLaMA hunts quants
Alibaba's Qwen 3.6 open weights for large-scale models remain MIA as the LocalLLaMA community scrambles for EXL3 quants. While the 35B-A3B variant is out, the flagship 72B release is clouded by rumors of internal shakeups.
The Qwen 3.6 release serves as a "vibe-check" for Alibaba's open-source commitment as internal turmoil threatens the roadmap for larger weights. While the 35B-A3B MoE variant excels in agentic coding, the flagship 72B release remains MIA, leaving the quant community deadlocked without official FP16 weights. Rumors of the Qwen open-source lead's departure have triggered "closed-source pivot" fears, especially as a massive capability gap grows between the proprietary 1M token "Plus" API and what can currently be run locally using hybrid offloading.
DISCOVERED
3h ago
2026-04-17
PUBLISHED
5h ago
2026-04-17
RELEVANCE
AUTHOR
2muchnet42day