Jundot Qwen3.8 Flash Next oQ4e MTP
A community mixed-precision MLX checkpoint of Qwen3.8-Flash-Next, including vision, n-gram embeddings and an MTP head, re-quantized on October 29.
What happened lately
Jundot Qwen3.8 Flash Next oQ4e MTP timeline
This is Jundot’s community quantization of Qwen’s Qwen3.8-Flash-Next, distributed in MLX format rather than GGUF. The English card lists the language backbone, n-gram embedding, vision encoder and multi-token prediction head, totaling 100.275 GiB of tensor storage according to its author.
The repository was created on August 27, 2026. An explicit October 29 card notice says the checkpoint was re-quantized with improved oQe in oMLX 0.7.3 and asks existing users to download it again. The card names Qwen Community License 1.0; this is not an Apache-licensed release or a new base model from Qwen.