🚨We just ran the initial set of benchmarks for #GLM 4.5 by Z.ai, and published the latest edition of our In Focus series comparing it to popular Chinese models such as DeepSeek V3 and MiniMax (official) M1. Read the full post here: layerlensai.substack.com/p/in-focus-glm…| Congrats to the

Latest on artificial intelligence developments thesequence.substack.com/p/the-sequence…

The Sequence Knowledge #697: The Most Important Theory in Modern AI Interpretability thesequence.substack.com/p/the-sequence…

The Sequence Knowledge #697: The Most Important Theory in Modern AI Interpretability: thesequence.substack.com/p/the-sequence…

Sometimes you eat the bear and sometimes the bear eats you. This is a very much a “you eat the bear” scenario. I was excited about Firelight before - this turbo charges it.

Running some math reasoning comparisons for the new OpenAI gpt-oss in LayerLens . In the AIME 2025 benchmark DeepSeek R1 performs a bit better than gpt-oss. However when compared to Qwen , gpt-oss really outperforms app.layerlens.ai

Some of the most important AI developments of last week: From GPT-5 to Claude Opus, This Crazy Week in Model Releases thesequence.substack.com/p/the-sequence…

looking forward to join Hugo Philion ☀️ tomorrow at The Wolf Of All Streets amazing podcast to chat about Firelight and all things XRPFi