Cheapest box to run Llama-4-Behemoth at Q3_K_M
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 Q3_K_M needs 981GB.
Nothing on the Grid runs Llama-4-Behemoth at Q3_K_M today. The largest Config, Mac Studio, M5 Ultra, 512GB, is 477GB 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−968GB headroom
- Mac mini, M6, 24GB$1,299Won't fit−960GB headroom
- Mac mini, M6, 32GB$1,499Won't fit−952GB headroom
- Mac mini, M5 Pro, 24GB$1,699Won't fit−960GB headroom
- Mac mini, M5 Pro, 48GB$2,499Won't fit−937GB headroom
- Mac Studio, M5 Max, 36GB$2,499Won't fit−948GB headroom
- Mac mini, M5 Pro, 64GB$2,899Won't fit−922GB headroom
- Mac Studio, M5 Max, 48GB$3,099Won't fit−937GB headroom
- Mac Studio, M5 Max, 64GB$3,499Won't fit−922GB headroom
- Won't fit−860GB headroom
- Won't fit−861GB headroom
- Won't fit−892GB headroom
- GeForce RTX 5090, 32GB$5,699.99+hostWon't fit−950GB headroom
- Won't fit−733GB headroom
- Won't fit−477GB headroom
| Config | Memory | Price | $/GB | Fit | Headroom | Est. speed |
|---|---|---|---|---|---|---|
| Mac mini, M6 | 16GB | $1,099 | $69 | Won't fit | −968GB | — |
| Mac mini, M6 | 24GB | $1,299 | $54 | Won't fit | −960GB | — |
| Mac mini, M6 | 32GB | $1,499 | $47 | Won't fit | −952GB | — |
| Mac mini, M5 Pro | 24GB | $1,699 | $71 | Won't fit | −960GB | — |
| Mac mini, M5 Pro | 48GB | $2,499 | $52 | Won't fit | −937GB | — |
| Mac Studio, M5 Max | 36GB | $2,499 | $69 | Won't fit | −948GB | — |
| Mac mini, M5 Pro | 64GB | $2,899 | $45 | Won't fit | −922GB | — |
| Mac Studio, M5 Max | 48GB | $3,099 | $65 | Won't fit | −937GB | — |
| Mac Studio, M5 Max | 64GB | $3,499 | $55 | Won't fit | −922GB | — |
| DGX Spark, GB10 Grace Blackwell | 128GB | $4,699 | $37 | Won't fit | −860GB | — |
| Mac Studio, M5 Max | 128GB | $5,099 | $40 | Won't fit | −861GB | — |
| Mac Studio, M5 Ultra | 96GB | $5,499 | $57 | Won't fit | −892GB | — |
| GeForce RTX 5090 | 32GB | $5,699.99+host | $178 | Won't fit | −950GB | — |
| Mac Studio, M5 Ultra | 256GB | $9,499 | $37 | Won't fit | −733GB | — |
| Mac Studio, M5 Ultra | 512GB | — | — | Won't fit | −477GB | — |
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 × 0.489 B/param (3.91 bits) + 0.250 GB/1K × 8K K/V + 1.2 GB runtime
speed = bandwidth × 0.5 ÷ (288B active × 0.489 B/param) = 140.8 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.