From the source
From the source
This is a weekly newsletter from Replicate covering open-source AI models, tools, and research.
It mentions DeepSeek-Coder-V2, an open-weights code model that briefly topped coding benchmarks, and PowerInfer-2, a fast inference engine for phones along with TurboSparse-Mistral-7B and TurboSparse-Mixtral-47B models.
It also discusses a research essay on AI search and tree search methods, and announces a new support bot feature in Discord.
From the source
The big news was DeepSeek-Coder-V2, the open-weights model that finally beat GPT-4o at coding tasks. It held the top spot on the leaderboard for a respectable four days before being beaten by the new Claude. The Mixture of Experts model comes in Full and Lite sizes and has a context length of 128K tokens.
replicate.com
