Cheapest box to run Llama-4-Behemoth at BF16
Meta’s Llama-4-Behemoth is 2000B parameters, 288B 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.
Llama-4-Behemoth at BF16 needs 4003GB.
Nothing on the Grid runs Llama-4-Behemoth at BF16 today. The largest Config, Mac Studio, M5 Ultra, 512GB, is 3499GB short.
Announced, weights not released. Parameter counts are Meta's; the K/V figure is an estimate from the Llama 4 family and is marked as one.
- Mac mini, M6, 16GB$1,099Won't fit−3990GB headroom
- Mac mini, M6, 24GB$1,299Won't fit−3982GB headroom
- Mac mini, M6, 32GB$1,499Won't fit−3974GB headroom
- Mac mini, M5 Pro, 24GB$1,699Won't fit−3982GB headroom
- Mac mini, M5 Pro, 48GB$2,499Won't fit−3959GB headroom
- Mac Studio, M5 Max, 36GB$2,499Won't fit−3970GB headroom
- Mac mini, M5 Pro, 64GB$2,899Won't fit−3944GB headroom
- Mac Studio, M5 Max, 48GB$3,099Won't fit−3959GB headroom
- Mac Studio, M5 Max, 64GB$3,499Won't fit−3944GB headroom
- Won't fit−3883GB headroom
- Won't fit−3883GB headroom
- Won't fit−3915GB headroom
- GeForce RTX 5090, 32GB$5,699.99+hostWon't fit−3972GB headroom
- Won't fit−3755GB headroom
- Won't fit−3499GB headroom
| Config | Memory | Price | $/GB | Fit | Headroom | Est. speed |
|---|---|---|---|---|---|---|
| Mac mini, M6 | 16GB | $1,099 | $69 | Won't fit | −3990GB | — |
| Mac mini, M6 | 24GB | $1,299 | $54 | Won't fit | −3982GB | — |
| Mac mini, M6 | 32GB | $1,499 | $47 | Won't fit | −3974GB | — |
| Mac mini, M5 Pro | 24GB | $1,699 | $71 | Won't fit | −3982GB | — |
| Mac mini, M5 Pro | 48GB | $2,499 | $52 | Won't fit | −3959GB | — |
| Mac Studio, M5 Max | 36GB | $2,499 | $69 | Won't fit | −3970GB | — |
| Mac mini, M5 Pro | 64GB | $2,899 | $45 | Won't fit | −3944GB | — |
| Mac Studio, M5 Max | 48GB | $3,099 | $65 | Won't fit | −3959GB | — |
| Mac Studio, M5 Max | 64GB | $3,499 | $55 | Won't fit | −3944GB | — |
| DGX Spark, GB10 Grace Blackwell | 128GB | $4,699 | $37 | Won't fit | −3883GB | — |
| Mac Studio, M5 Max | 128GB | $5,099 | $40 | Won't fit | −3883GB | — |
| Mac Studio, M5 Ultra | 96GB | $5,499 | $57 | Won't fit | −3915GB | — |
| GeForce RTX 5090 | 32GB | $5,699.99+host | $178 | Won't fit | −3972GB | — |
| Mac Studio, M5 Ultra | 256GB | $9,499 | $37 | Won't fit | −3755GB | — |
| Mac Studio, M5 Ultra | 512GB | — | — | Won't fit | −3499GB | — |
GeForce RTX 5090 is a card, not a computer: its price is the US street median plus $1,000 for the minimum host machine. NVIDIA's $1,999 MSRP is not a price anyone pays.
memory = 2000B × 2.000 B/param (16 bits) + 0.250 GB/1K × 8K K/V + 1.2 GB runtime
speed = bandwidth × 0.5 ÷ (288B active × 2.000 B/param) = 576.0 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.