MagicAI claims pretraining recipe matches frontier model with 50x less compute, ~$0.5M
EricSteinb · x · 2026-09-09
MagicAI Labs is pushing back on the idea that frontier pretraining is a big-lab-only game: without 100k chips, their answer is algorithmic efficiency.
- Claims its new recipe matches DeepSeek V4 Pro's pretraining using 50x less compute
- Roughly half the FLOPs used for GPT-3
- About $0.5M on GB200 hardware
Caveat: this is a self-reported claim with no independent verification yet.
Related event: Magic Claims to Match DeepSeek V4 Pro Pretraining with 1/50 the Compute(11 posts)→
More from Infra
- Anthropic's ~11 GW contracted compute estimated at ~42% NVIDIA GPUs, ~4.5 GW Google TPU — ns123abc · 2026-10-27
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- RTX 5080 price jumps from $1658 to $2318 in ten days on Amazon — Frizzy-MacDrizzle · 2026-10-27
- ComfyUI on 12GB VRAM: measured peaks show offloading turns 24s into 11 minutes — Impossible-Mail-4619 · 2026-10-27
- Neon rebuilds its docs search entirely on its own stack: Lakebase Search, Functions, AI Gateway — matei_zaharia · 2026-10-27
- PyTorch details 2.12x faster on-device streaming ASR on Arm SME2 via ExecuTorch — PyTorch · 2026-10-27