Cheapest box to run Llama-4-Behemoth at Q8_0
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 Q8_0 needs 2128GB.
Nothing on the Grid runs Llama-4-Behemoth at Q8_0 today. The largest Config, Mac Studio, M5 Ultra, 512GB, is 1624GB 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−2115GB headroom
- Mac mini, M6, 24GB$1,299Won't fit−2107GB headroom
- Mac mini, M6, 32GB$1,499Won't fit−2099GB headroom
- Mac mini, M5 Pro, 24GB$1,699Won't fit−2107GB headroom
- Mac mini, M5 Pro, 48GB$2,499Won't fit−2084GB headroom
- Mac Studio, M5 Max, 36GB$2,499Won't fit−2095GB headroom
- Mac mini, M5 Pro, 64GB$2,899Won't fit−2069GB headroom
- Mac Studio, M5 Max, 48GB$3,099Won't fit−2084GB headroom
- Mac Studio, M5 Max, 64GB$3,499Won't fit−2069GB headroom
- Won't fit−2008GB headroom
- Won't fit−2008GB headroom
- Won't fit−2040GB headroom
- GeForce RTX 5090, 32GB$5,699.99+hostWon't fit−2097GB headroom
- Won't fit−1880GB headroom
- Won't fit−1624GB headroom
| Config | Memory | Price | $/GB | Fit | Headroom | Est. speed |
|---|---|---|---|---|---|---|
| Mac mini, M6 | 16GB | $1,099 | $69 | Won't fit | −2115GB | — |
| Mac mini, M6 | 24GB | $1,299 | $54 | Won't fit | −2107GB | — |
| Mac mini, M6 | 32GB | $1,499 | $47 | Won't fit | −2099GB | — |
| Mac mini, M5 Pro | 24GB | $1,699 | $71 | Won't fit | −2107GB | — |
| Mac mini, M5 Pro | 48GB | $2,499 | $52 | Won't fit | −2084GB | — |
| Mac Studio, M5 Max | 36GB | $2,499 | $69 | Won't fit | −2095GB | — |
| Mac mini, M5 Pro | 64GB | $2,899 | $45 | Won't fit | −2069GB | — |
| Mac Studio, M5 Max | 48GB | $3,099 | $65 | Won't fit | −2084GB | — |
| Mac Studio, M5 Max | 64GB | $3,499 | $55 | Won't fit | −2069GB | — |
| DGX Spark, GB10 Grace Blackwell | 128GB | $4,699 | $37 | Won't fit | −2008GB | — |
| Mac Studio, M5 Max | 128GB | $5,099 | $40 | Won't fit | −2008GB | — |
| Mac Studio, M5 Ultra | 96GB | $5,499 | $57 | Won't fit | −2040GB | — |
| GeForce RTX 5090 | 32GB | $5,699.99+host | $178 | Won't fit | −2097GB | — |
| Mac Studio, M5 Ultra | 256GB | $9,499 | $37 | Won't fit | −1880GB | — |
| Mac Studio, M5 Ultra | 512GB | — | — | Won't fit | −1624GB | — |
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 × 1.063 B/param (8.5 bits) + 0.250 GB/1K × 8K K/V + 1.2 GB runtime
speed = bandwidth × 0.5 ÷ (288B active × 1.063 B/param) = 306.0 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.