Stanford's Marin 535B Model Training Starts with Full Transparency
udmrzn · x · 2026-08-24
Stanford professor Percy Liang announced the start of training for the new Marin 535B-A23B model, opening the entire process to the public. The model features 535B total parameters (23B active), trained on 18.75T tokens using 11 GB200 NVL72 clusters (approx. 792 GPUs) for an estimated 3 months and 2.7e24 FLOPs. Prior to the main run, the team completed a 4-rung Scaling Ladder (from 1.6B to 27.7B) to debug and forecast training behavior. The public can now monitor real-time training curves and logs, marking this as potentially the largest openly observed model training run to date.
More from Infra
- engy.ai processes 50B+ tokens in a day as open-model inference demand surges — const_reborn · 2026-10-16
- New inference engine 'e' hits 10,500 tok/s on a single DGX Station — brandon_galang · 2026-10-16
- SGLang ships Day-0 support for FLUX 3 Image Klein: 10 reference images, BF16/FP8/NVFP4 variants — ying11231 · 2026-10-16
- NVIDIA's Nemotron 3.5 Super VL hits CoreWeave: 12B-active multimodal MoE at $0.25/M input tokens — wandb · 2026-10-16
- Chip equipment beats bull case: 2026 WFE seen at $153B, $278B by 2030 — BenBajarin · 2026-10-16
- Steam now launches in 5.23s, beating Epic's 5.47s — and vLLM gets roasted for slow startup — AlpinDale · 2026-10-16