I've seen a number of claims that it's 10T or 6T paramètres and MoE, but without sources. Simon Willison has some good points about why it certainly feels big: simonwillison.net/2026/Jun/9/c...
Who's Who of AI
Jelle Zuidema 🟥
Associate Professor of Natural Language Processing & Explainable AI, University of Amsterdam, ILLC
What they're sharing
Initial impressions of Claude Fable 5 simonwillison.net
Tracking the emergence of linguistic structure in self-supervised models learning from speech arxiv.org
Articles & links
Jelle Zuidema 🟥 reposted
Let’s study learning trajectories in self-supervised speech models! 🔊 Do they reflect the hierarchical organization of spoken language? We have analyzed a lot of training checkpoints to find out 🌠 Preprint: arxiv.org/abs/2604.02043 ⬇️
AI Weekly's analysis
→
- Researchers probed six Wav2Vec2 and HuBERT models trained on spoken Dutch across layers and intermediate training checkpoints.
- Different levels of linguistic structure showed distinct layerwise patterns and learning trajectories inside the same speech models.
- Higher-order prediction tasks using iteratively refined pseudo-labels induced greater parallelism across the models' internal layers.
Read full analysis →
View on Bluesky →
Their own posts
Recent commentary
Does anyone know of some plausible speculations on *technical innovations* driving the impressive performance of Mythos/Fable? The model card only talks about evaluations (interestingly, mostly in biology). The interpretability work on Mythos Preview suggests it's essentially all based on versions
Their network
In Jelle Zuidema 🟥's orbit
Center = Jelle Zuidema 🟥. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.