ComputeAssay

Model guide · prices from 2026-09-10 · API prices daily, full survey weekly

Cheapest way to run DeepSeek-R1 (671B MoE)

DeepSeek-R1 (671B MoE) needs ≈403 GB of VRAM at 4-bit — this week that starts at $5.21/hour (13× RTX5090, cheapest tracked walk-up price). FP8 and FP16 below. VRAM calculators stop at the memory number; this connects it to live, graded prices across 20 providers.

MoE note: DeepSeek-R1 (671B MoE) activates ~37B active per token, but the full 671B weights must fit in VRAM — memory scales with total parameters, speed with active ones.

FP16 — ≈1610 GB needed

GPUVRAMGPUs neededCheapest trackedVerifiable A/B
A100-SXM80 GB21 (multi-node)$19.05/hr$48.30/hr
RTX509032 GB51 (multi-node)$20.45/hr
RTX409024 GB68 (multi-node)$25.43/hr
A100-PCIe80 GB21 (multi-node)$28.35/hr
MI300X192 GB9 (multi-node)$31.05/hr
H100-PCIe80 GB21 (multi-node)$41.79/hr
B300288 GB6$45.00/hr$106.81/hr
L40S48 GB34 (multi-node)$46.58/hr

FP8/INT8 — ≈805 GB needed

GPUVRAMGPUs neededCheapest trackedVerifiable A/B
A100-SXM80 GB11 (multi-node)$9.98/hr$25.30/hr
RTX509032 GB26 (multi-node)$10.43/hr
RTX409024 GB34 (multi-node)$12.72/hr
A100-PCIe80 GB11 (multi-node)$14.85/hr
MI300X192 GB5$17.25/hr
H100-PCIe80 GB11 (multi-node)$21.89/hr
B300288 GB3$22.50/hr$53.41/hr
L40S48 GB17 (multi-node)$23.29/hr

4-bit — ≈403 GB needed

GPUVRAMGPUs neededCheapest trackedVerifiable A/B
RTX509032 GB13 (multi-node)$5.21/hr
A100-SXM80 GB6$5.44/hr$13.80/hr
RTX409024 GB17 (multi-node)$6.36/hr
A100-PCIe80 GB6$8.10/hr
MI300X192 GB3$10.35/hr
H100-PCIe80 GB6$11.94/hr
H200-SXM141 GB3$11.97/hr$12.87/hr
L40S48 GB9 (multi-node)$12.33/hr

Method: VRAM ≈ parameters × bytes/parameter × 1.2 (20% headroom for KV cache and activations at ~8K context). Longer context or high concurrency needs more; training needs far more (optimizer states: ~4–8× weights). GPU counts assume even sharding. Above one 8-GPU node, the interconnect becomes the constraint — restrict to Grade A/B listings. Grades measure what sellers publish — deliverability of the written claim, not measured performance. No fleet holds a measured (Verified) badge yet; that tier arrives with the verification harness. Prices are this week's cheapest tracked walk-up listings (teaser prices excluded); every listing with grades and sources on the GPU pages. Interactive version for any model size: Model Fit.

These prices move weekly

The Delivered Compute Report tracks them — free, every row sourced.

Subscribe free