@SemiAnalysis_
SITUATION DETECTED: The city of Rio de Janerio has post-trained a model. Based on Qwen 7/2, Rio 3.5 Open 397B adds SwiReasoning on top of the base Qwen model — a framework that dynamically switches between standard chain-of-thought and latent-space reasoning, guided by entropy-based confidence signals, so the model only "thinks out loud" when it needs to and otherwise reasons silently in hidden space for better token efficiency.
DAY 0 ALERT: @MiniMax_AI M3 is now available on HuggingFace & has been added to InferenceX. The M3 architecture has ~428B parameters and ~23B activated parameters. Due to the 10x engineers from @inferact, M3 is already delivering pretty well-optimized performance on @NVIDIAAI B300 Blackwell Ultra on Day 0 @vllm_project! Furthermore, Inferact released their EAGLE3 heads, which enable even greater performance. Looking forward to Day 1, 2, and 3 performance & the team is grinding on benchmarking Day 0 MI355X performance on InferenceX too.
we heard fable got banned
Congrats to @vllm_project & @lmsysorg for releasing MiniMax M3 428B on both the CUDA & ROCm stack on day 0! MiniMax M3 includes: 🟠 Block sparse attention which is 9x faster prefill over M2.7 🟠 Day 0 open MXFP8 weights 🟠 and Furthermore @Inferact released Day-0 EAGLE3 open weight draft model support Excited to try out the performance on MiniMax M3!
Morgan Stanley ECM on $SPCX: "engaging stabilizers..." 🚀
The concept of “80,000 hours” career consulting doesn’t even make sense. If someone wants to have a high-impact life, they would be working more than 80,000 hours, i.e. more than 40 hours a week. They should rename themselves to 160,000 Hours. If you want to have a high-impact career and are a motivated AI engineer, email us: letsgo@SemiAnalysis.com
Interestingly, the public market is positioned in the opposite direction, with neocloud names trading like the cycle is about to roll over. Our read, which we lay out in the piece, is that the scarcity is real, the long-dated rental floor is much higher than the equity setup implies, and existing H100 fleets have meaningfully more economic life left than the consensus model assumes. Link to the Newsletter: (4/4) t.co/3AqjN4T5nk
