Alibaba Launches Qwen 3.5 Models as OpenAI Military Deal Drives Claude to #1
Large Language Models
Alibaba Releases Qwen 3.5 Model Series and Achieves High Local Performance: Alibaba has launched a new series of Qwen 3.5 models, including 0.8B, 2B, 4B, and 9B variants, designed to meet diverse computational requirements. Community reports highlight that larger versions, like the 27B dense model, can now achieve significant speeds on consumer-grade hardware, reflecting rapid advancements in local LLM efficiency over the past year.
- Breaking : Today Qwen 3.5 small
- Running Qwen3.5 27b dense with 170k context at 100+t/s decode and ~1500t/s prefill on 2x3090 (with 585t/s throughput for 8 simultaneous requests)
- 13 months since the DeepSeek moment, how far have we gone running models locally?
Jan Team Releases Jan-Code-4B: The Jan team has introduced Jan-code-4B, a specialized model fine-tuned for coding tasks such as debugging, generation, and test writing. This small-scale model is intended to be a drop-in replacement for the Haiku model in Claude Code, offering improved performance over its baseline version.
AI Tools & Platforms
Alibaba Open-Sources CoPaw Agent Workstation: Alibaba has open-sourced CoPaw, a high-performance personal workstation designed to help developers manage multi-channel AI workflows and complex memory. The tool focuses on enhancing productivity by scaling the management of sophisticated AI agent tasks.
Industry Ethics & Market Trends
OpenAI Military Deal Triggers User Migration to Claude: A new partnership between OpenAI and the US military has sparked a "cancel ChatGPT" movement, leading to a significant migration of users to competitors. Anthropic’s Claude app reached No. 1 on the App Store as users sought alternatives that align with more restrictive stances on military AI applications.