@latkins: Little late notice but I’ll be speaking here in 30 minutes. An updated version of my Trinity Large talk with some tease…
Summary
Alatkins announces a last-minute talk at AI4 Conference in Vegas, covering an updated version of the Trinity Large talk with teasers about training a 400B MoE model to 17T tokens without loss spikes.
View Cached Full Text
Cached at: 08/05/26, 10:29 PM
Little late notice but I’ll be speaking here in 30 minutes. An updated version of my Trinity Large talk with some teasers of what’s to come.
Arcee.ai (@arcee_ai): If you’re in Vegas for @Ai4Conferences, be sure to stop by @latkins’ 11:20 AM session in Venetian Ballroom A:
How not to blow up: Training a 400B MoE to 17T tokens without loss spikes
Similar Articles
@PyTorch: How do we get more useful work—not just more tokens—from every AI dollar? This Wednesday at 11:50 AM at @AMD #Advancing…
PyTorch Foundation CTO Matt White will speak at AMD's AdvancingAI event about optimizing AI inference economics using open-source tools like vLLM and SGLang, advocating for right-sized models and intelligent routing to improve dollar per intelligence.
@gmkurtzer: I just listened to this video from @latkins who is CTO of @arcee_ai talking about lessons learns of training a large MO…
Gregory Kurtzer praises Lucas Atkins' talk on lessons learned from training a large sparse Mixture-of-Experts model and life at an AI lab startup.
Liquid AI reveals 8B-A1B MoE trained on 38T
Liquid AI released LFM2.5-8B-A1B, an edge MoE model trained on 38T tokens with a 128K context window, improved tool calling, and reasoning capabilities, available on Hugging Face.
@Prince_Canuma: My @aiDotEngineer talk is live: "On-device Intelligence using MLX" Huge thanks to @swyx and the team for having me — ha…
The author announces their live talk titled 'On-device Intelligence using MLX' at the aiDotEngineer event, expressing gratitude to the organizers and community contributors.
@svlevine: Starting now in ASEM Ballroom 203 after a bit of AV troubles :) see you there! The talk will cover how to build up a ge…
Sergey Levine announces his talk at the ICML 2026 Workshop on Decision-Making from Offline Datasets, covering foundations of data-driven decision making from model-based optimization to offline RL.