Vercel has added Reflection AI's Beam, the model previously tested under the stealth name Cat Jam, to its AI Gateway, free for a limited launch period. Beam is a US open-weight reasoning model under Apache 2.0, aimed at coding and agent work, with a 1M-token context and only 23B of its 501B parameters active per token.
Hugging Face is hosting Beam, an open model with 501 billion parameters. The release is pitched as a model that users can own, adapt to their needs and run at scale.
IBM's Granite team presents Granite Switch 4.1 on Hugging Face, which brings chat, decision-making and speech together behind a single endpoint.
Google opened Gemini 4 Argon to developers and businesses worldwide on October 13, at a launch price of $2 per million input tokens. That rate is set to double on December 1, raising costs for teams building on the model.
Liquid AI's d1-ultra takes text and images and returns calibrated probabilities in a single forward pass. It scores 65 on Artificial Analysis's Decision Index and is priced at $0.042 per million input tokens.
Meta has launched Muse Spark Open, an open model with 816 billion parameters, 42 billion of them active. Meta positions it as a response to the open-weight models coming out of China.