#cohere
cohere
519 views
2 Discussing
Hot
Latest
Cohere drops its first open-source agent programming model, North Mini Code, focusing on sovereign AI and high throughput.
Cohere announced the open-source release of its first large model for agent programming, North Mini Code, built on the MoE architecture with a total parameter count of 30B, activating only 3B parameters during each forward pass. The model is available under the Apache 2.0 license, allowing deployment on local or private clouds.
In programming benchmarks, North Mini Code scored 33.4, with an output throughput 2.8 times that of similar models, and a 30% reduction in latency between tokens. The model is specifically trained for Agent workflows, excelling at coordinating sub-agents, drafting architecture diagrams, and conducting code reviews.
Why it matters: This is a significant breakthrough for Cohere in the AI programming space, and the open-source strategy will drive the decentralization of AI development tools, reducing reliance on cloud vendors for enterprises.
#Cohere #AI #开源 #Agent #Web3
Cohere announced the open-source release of its first large model for agent programming, North Mini Code, built on the MoE architecture with a total parameter count of 30B, activating only 3B parameters during each forward pass. The model is available under the Apache 2.0 license, allowing deployment on local or private clouds.
In programming benchmarks, North Mini Code scored 33.4, with an output throughput 2.8 times that of similar models, and a 30% reduction in latency between tokens. The model is specifically trained for Agent workflows, excelling at coordinating sub-agents, drafting architecture diagrams, and conducting code reviews.
Why it matters: This is a significant breakthrough for Cohere in the AI programming space, and the open-source strategy will drive the decentralization of AI development tools, reducing reliance on cloud vendors for enterprises.
#Cohere #AI #开源 #Agent #Web3
Log in to explore more content