Stanford Marin releases all 57 intermediate checkpoints of Mochi 535B-A23B, 18.2T tokens trained
saurabh_shah2 · x · 2026-11-18
dlwh announced that the Marin project is releasing all 57 intermediate checkpoints from the "Mochi" 535B-A23B pretraining run, including 10 from long-context extension — 18.2T tokens trained, roughly 61 TB of weights, with an accompanying blog post.
Reposters shared it as "new model organism just dropped," with a relay-baton meme of Marin handing off from OLMO — full transparency into the pretraining process (not just the final model) continues the OLMO-style open research tradition, letting the community study training dynamics and scaling behavior.
Related event: Stanford Marin Releases All 57 Intermediate Checkpoints of 535B MoE Model(5 posts)→
More from Research
- sokrypton predicts protein structures from LLM contact maps, admits the structure part is a gimmick — sokrypton · 2026-11-18
- OpenAI study: persona-style CoT training drops monitorability from 0.94 to 0.22 — dair_ai · 2026-11-18
- LLM protein-folding test on fresh PDB entry lands near-random structures — MoAlQuraishi · 2026-11-18
- Lean-PEP turns worst-case optimization bounds into formally verified Lean proofs — prof_grimmer · 2026-11-18
- Two arms, two token streams, one shared vocabulary: new robot manipulation work open-sourced — DJiafei · 2026-11-18
- NUS's BiCAT action tokenizer for bimanual robots: 113/150 real-robot wins, 2.7x lower latency than OAT — DJiafei · 2026-11-18