Ai2's tech report shows how 512 GPUs train one model—and flags 'token gerrymandering'
allen_ai · x · 2026-10-01
Alongside Olmo-core 3, Ai2 published a tech report and an interactive explainer walking through how 512 GPUs train a single large model: from single-GPU compute/memory basics through data parallelism, memory sharding, and expert parallelism. The report also details experiments and failures, including 'token gerrymandering'—a case where a load-balancing score improved even as workloads became less balanced, showing the metric can be misleading.
Related event: Ai2 Open-Sources Olmo-core 3, a Trillion-Parameter MoE Training Stack(5 posts)→
More from Infra
- Gemma.c: full Gemma 1/2/3 multimodal inference in a single C file, ~72 tok/s on CPU at INT8 — This-Purchase-3325 · 2026-11-16
- Sarvam AI Launches Token Factory: India-Hosted Inference for Open Models — itsOmSarraf_ · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16
- vLLM serves DeepSeek V4.1-Flash at 954K tokens/s per GPU, $38.96/GPU-hour on GB300 NVL72 — bookwormengr · 2026-11-16