gemma-3-27b-it (google) is a 27.4B parameters open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 21.55 GB VRAM for Q4 (weights 15.07 GB + KV 5.48 GB + 1 GB overhead) and 61.28 GB for FP16. About 28 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 185.8 tok/s (estimated); MI300X 192GB ≈ 123.1 tok/s (estimated); H200 141GB SXM5 ≈ 111.5 tok/s (estimated); H100 80GB SXM5 ≈ 77.8 tok/s (estimated); Gaudi 3 128GB ≈ 76.1 tok/s (estimated). Use the table below for VRAM fit and tok/s per dollar (list MSRP). Estimates follow memory-bandwidth math; measured rows override when present. See methodology for details. Methodology.