2️⃣ Second, we wanted to test it against models tailored for coding tasks. Unsurprisingly, it performs way under those. Llama 4 Maverick achieved only a 70% accuracy score. Alibaba’s Qwen2.5-Coder-32B is ranking the best at (90%), closely followed by GPT o3-mini (89%).

1️⃣ First, we wanted to reproduce Meta's findings that Llama 4 outperformed GPT-4o, Gemini 2.0 Flash, and DeepSeek v3.1—we found the exact opposite.

It came last, 6% less than the next best-performing model (DeepSeek) and 18% behind the overall top-performing model (GPT-4o).

Just finished building @rootly.com MCP server: go from incident to resolution in under a minute. ⏱️

-Plug it into your IDE
-Import an incident in Cursor’s chat
-Cursors investigate the issue based on the metadata
-Cursors suggest a fix, review, and save

github.com/Rootly-AI-La...

Obviously, @MistralAI promoted how good Le Chat is at finding food pairings for wine 🇫🇷.

That should be included in all model benchmarks.

Perses, a sandboxed @cncf.bsky.social porject, provides standards for visualization and dashboards for metrics monitoring.

@schabell.org is sharing everything you need to know about the project

Looking for where to store your AI assets? Harbor - the @cncf.bsky.social incubating project - might be what you are looking for.

Learn more from Harbor maintainer Vadim Bauer by watching the full episode 👇

View from Hawaii Diamond Head