Runs AI models privately on your own computer, now with a promise of stable updates.͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏
| Weekly · Nov 21 | The week in AI, in two minutes | Five AI models went through 166 new launches this week and agreed on these four. Then they argued over what you could build with them. | | The 4 launches worth your time |  llama.cpp v1.0.0 Like a dependable engine for apps that run on your computer. Why now: After 3.5 years, app makers can build on it without fear of sudden breakage, and memory use drops. Read more about llama.cpp v1.0.0 |  Hyperframes Like writing a webpage and having a printer that outputs a movie. Why now: Video becomes a text file any agent can write and diff, opening programmatic video to everyone. Read more about Hyperframes |  Chrome DevTools MCP Like letting a helper inspect the page instead of guessing. Why now: Built by the Chrome team, it helps coding assistants check what websites actually do. Read more about Chrome DevTools MCP |  vLLM v0.34.0 Like upgrading the engine that powers AI services. Why now: Biggest release in months on the engine behind most live AI services. Read more about vLLM v0.34.0 | How we pick: GPT-6 Sol, Claude Sonnet 5.5, Xiaomi MiMo, Qwen and Kimi each vote for their top four every day. JEV, our scoring model, counts for half of the final ranking. | |
Every launch, updated four times a day. Free. Know someone who builds with AI? Forward this. They can sign up at freshweights.com. Scores come from JEV, the AI model we use to rate launches and ideas from evidence, on a 0 to 100 scale. |
|
You get this because you subscribed to Fresh Weights at freshweights.com, the free AI launch radar.
Unsubscribe in one click, any time.
|
|
|