[AINews] Qwen 3.8 Max(2.4T) and 27B, new open weights models for Coding and Cowork
Alibaba says Qwen3.8-Max will move from API access to open weights, with Qwen3.8-27B following it.
Latent Space frames the 2.4T-parameter Max model as a major return to open-weight releases for Qwen after doubts about the lab’s direction. The note cites Alibaba’s pricing at $2 per million input tokens and $6 per million output tokens, plus claims around long-horizon coding, chip-design work, e-commerce simulations, and multimodal agents. It also flags the practical caveat: a 2.4T-class model may be open-weight without being easy to self-host. Licensing concerns around geographic restrictions were raised in the discussion but not resolved in the piece. Latent Space's note
Latent Space frames the 2.4T-parameter Max model as a major return to open-weight releases for Qwen after doubts about the lab’s direction. The note cites Alibaba’s pricing at $2 per million input tokens and $6 per million output tokens, plus claims around long-horizon coding, chip-design work, e-commerce simulations, and multimodal agents. It also flags the practical caveat: a 2.4T-class model may be open-weight without being easy to self-host. Licensing concerns around geographic restrictions were raised in the discussion but not resolved in the piece. Latent Space's note
score 8