Cheapest box to run Kimi-K2-Instruct at BF16
Moonshot’s Kimi-K2-Instruct is 1000B 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.
Kimi-K2-Instruct at BF16 needs 2002GB.
Nothing on the Grid runs Kimi-K2-Instruct at BF16 today. The largest Config, Mac Studio, M5 Ultra, 512GB, is 1498GB short. At it fits with 13GB headroom.
A trillion parameters. No Config we track holds it at any quant, and we list it so the arithmetic says so plainly.
- Mac mini, M6, 16GB$1,099Won't fit−1989GB headroom
- Mac mini, M6, 24GB$1,299Won't fit−1981GB headroom
- Mac mini, M6, 32GB$1,499Won't fit−1973GB headroom
- Mac mini, M5 Pro, 24GB$1,699Won't fit−1981GB headroom
- Mac mini, M5 Pro, 48GB$2,499Won't fit−1958GB headroom
- Mac Studio, M5 Max, 36GB$2,499Won't fit−1969GB headroom
- Mac mini, M5 Pro, 64GB$2,899Won't fit−1943GB headroom
- GeForce RTX 5090, 32GB$2,999.99+hostWon't fit−1971GB headroom
- Won't fit−1943GB headroom
- Mac Studio, M5 Max, 48GB$3,099Won't fit−1958GB headroom
- Mac Studio, M5 Max, 64GB$3,499Won't fit−1943GB headroom
- Won't fit−1881GB headroom
- Won't fit−1913GB headroom
- Won't fit−1882GB headroom
- Won't fit−1881GB headroom
- Won't fit−1882GB headroom
- Won't fit−1913GB headroom
- Won't fit−1754GB headroom
- Won't fit−1754GB headroom
- Won't fit−1498GB headroom
| Config | Memory | Price | $/GB | Fit | Headroom | Est. speed |
|---|---|---|---|---|---|---|
| Mac mini, M6 | 16GB | $1,099 | $69 | Won't fit | −1989GB | — |
| Mac mini, M6 | 24GB | $1,299 | $54 | Won't fit | −1981GB | — |
| Mac mini, M6 | 32GB | $1,499 | $47 | Won't fit | −1973GB | — |
| Mac mini, M5 Pro | 24GB | $1,699 | $71 | Won't fit | −1981GB | — |
| Mac mini, M5 Pro | 48GB | $2,499 | $52 | Won't fit | −1958GB | — |
| Mac Studio, M5 Max | 36GB | $2,499 | $69 | Won't fit | −1969GB | — |
| Mac mini, M5 Pro | 64GB | $2,899 | $45 | Won't fit | −1943GB | — |
| GeForce RTX 5090 | 32GB | $2,999.99+host | $94 | Won't fit | −1971GB | — |
| Mac Studio, M4 Max, refurb | 64GB | $3,049 | $48 | Won't fit | −1943GB | — |
| Mac Studio, M5 Max | 48GB | $3,099 | $65 | Won't fit | −1958GB | — |
| Mac Studio, M5 Max | 64GB | $3,499 | $55 | Won't fit | −1943GB | — |
| Ascent GX10, GB10 Grace Blackwell | 128GB | $3,999.99 | $31 | Won't fit | −1881GB | — |
| Mac Studio, M3 Ultra, refurb | 96GB | $4,329 | $45 | Won't fit | −1913GB | — |
| Mac Studio, M4 Max, refurb | 128GB | $4,409 | $34 | Won't fit | −1882GB | — |
| DGX Spark, GB10 Grace Blackwell | 128GB | $4,699 | $37 | Won't fit | −1881GB | — |
| Mac Studio, M5 Max | 128GB | $5,099 | $40 | Won't fit | −1882GB | — |
| Mac Studio, M5 Ultra | 96GB | $5,499 | $57 | Won't fit | −1913GB | — |
| Mac Studio, M3 Ultra, refurb | 256GB | $7,729 | $30 | Won't fit | −1754GB | — |
| Mac Studio, M5 Ultra | 256GB | $9,499 | $37 | Won't fit | −1754GB | — |
| Mac Studio, M5 Ultra | 512GB | — | — | Won't fit | −1498GB | — |
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 = 1000B × 2.000 B/param (16 bits) + 0.070 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.