• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
AI

Marin 535B Model Starts Open Pretraining on 18.75T Tokens

Researchers begin transparent pretraining of a 535B-parameter model using public data and configs.

Miles BrundageMB
Dan RoyDR
Alex DimakisAD
17 Sources, 50d ago, first seen 50d ago

TLDR

Posts from researchers and engineers report that the Marin project has started pretraining its 535B-parameter model on 18.75 trillion tokens. The run uses 11 NVIDIA GB200 NVL72 systems and is projected to last roughly three months. Participants describe the effort as fully open, with public pretraining mixtures, sampled documents, live loss curves, and training configs. Replies highlight the scale and transparency while comparing it to prior open initiatives. No independent verification of completion or final results appears in the packet.

Combined views

109K

17 Sources, first seen 50d ago

1.5K likes23 comments663 saves172 reposts

Sources

  1. EL
    elie@eliebakouch7 weeks ago

    marin is the most open model training initiative in the world you can visualize the exact pre training mixture per domain, look at sampled documents, follow the live training loss, dig into configs, scaling laws https://twitter.com/percyliang/status/2090918065634684997

    • likes: 860
    • replies: 12
    • bookmarks: 491
    • reposts: 59
  2. RI
    rishi@RishiBommasani7 weeks ago

    5 years ago, almost to the day, @siddkaramcheti @laurel_orr1 @percyliang put out their first public writing on training open models. https://crfm.stanford.edu/2021/08/26/mistral.html Even then, model training was not for the faint of heart. Mistral (before there was @MistralAI)…

    • likes: 43
    • replies: 0
    • bookmarks: 22
    • reposts: 7
  3. JA
    Julius Adebayo@juliusadml7 weeks ago

    pre-registering your pretraining loss forecast is the ultimate confidence move. 535B moe training happening all in the open. https://twitter.com/percyliang/status/2090918067987689486

    • likes: 15
    • replies: 0
    • bookmarks: 6
    • reposts: 0
  4. AD
    Alex Dimakis@AlexGDimakis7 weeks ago

    In medieval times, technical knowledge was kept hidden : guilds guarded recipes, processes, and tools and alchemists often wrote in deliberately obscure language. Knowledge was kept secret because it gives economic, professional, or political advantage. Enlightenment and the…

    • likes: 405
    • replies: 10
    • bookmarks: 123
    • reposts: 44
  5. ZW
    Zhaoran Wang@zhaoran_wang7 weeks ago

    well said! https://twitter.com/AlexGDimakis/status/2091021829079273921

    • likes: 2
    • replies: 0
    • bookmarks: 0
    • reposts: 0
  6. AK
    Amin Karbasi@aminkarbasi7 weeks ago

    Open science AI in its finest form. https://twitter.com/percyliang/status/2090918065634684997

    • likes: 13
    • replies: 0
    • bookmarks: 2
    • reposts: 1

Combined views

109K

17 Sources, first seen 50d ago

1.5K likes23 comments663 saves172 reposts

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

MarinPercy LiangStanford

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Featured Source
PLPercy Liang@percyliang2:44 PM · Aug 21, 2026

🚢 Marin 535B-A23B started training this week! As usual, the whole process is open. Voyage plan: pretraining (80%) + midtraining (20%) on 18.75T tokens on 11 x GB200 NVL72 for ~3 months (2.7e24 FLOPs). Post-training will follow. Before kicking off the run, we trained a 4-rung…

    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet