Local AI Deals

Cheapest box to run GLM-4.6 at BF16

Z.ai’s GLM-4.6 is 357B parameters, 32B of them read per token. The number below is computed from that and each vendor’s published specs, not asserted. Switch the quant or the context and it recomputes in place.

Fit finderspecs, not prices
Quant
Context

GLM-4.6 at BF16 needs 718GB.

Nothing on the Grid runs GLM-4.6 at BF16 today. The largest Config, Mac Studio, M5 Ultra, 512GB, is 214GB short. At it fits with 69GB headroom.

GeForce RTX 5090 is a card, not a computer: every price shown for it is the cheapest listing we found plus $1,000 for the minimum host machine.

memory = 357B × 2.000 B/param (16 bits) + 0.377 GB/1K × 8K K/V + 1.2 GB runtime
speed = bandwidth × 0.5 ÷ (32B active × 2.000 B/param) = 64.0 GB read per token
usable = memory − operating system reserve (8% on a Mac, 6% on a DGX Spark, 4% on a card)

Bandwidth is the vendor’s published figure. Every speed is an estimate and says so. Methodology.