Iota says Orion-16B is training live on unreliable compute across three continents
markjeffrey · x · 2026-07-28
Iota says it is training a 16B model on unreliable, distributed compute across three continents
Macrocosmos AI says a 16-billion-parameter model, Orion-16B, is now training live on IOTA’s distributed network.
- The setup spans three continents and uses heterogeneous, permissionless compute owned by no single entity.
- The key engineering claim is resilience: training computers can drop in and out, and the run keeps going.
- The post frames this as the next stage of Project Orion, with more details promised on the systems, providers, and engineering behind the training stack.
Related event: Project Orion Trains 16B Model Across Three Continents(3 posts)→
More from Infra
- AMD opens Instella-MoE, a fully open 16B MoE model trained on MI300X and MI325X — QuixiAI · 2026-07-28
- Half of U.S. servers may sit in tiny rooms, not giant Virginia data centers — aronchick · 2026-07-28
- Morgan Stanley says AI memory prices may peak in Q4 2026 as NAND inventories rise — SumitGup · 2026-07-28
- Databricks pitches AI spend governance as token usage outpaces value — matei_zaharia · 2026-07-28
- A company with 2 H200s asks which coding model and vLLM setup can serve 4–10 users — redblood252 · 2026-07-28
- AMD, Intel and chip-equipment demand headline The Circuit’s latest AI infrastructure episode — BenBajarin · 2026-07-28