Cheapest box to run Kimi-K2-Instruct at Q3_K_M
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 Q3_K_M needs 491GB.
Cheapest Config that fits: Mac Studio, M5 Ultra, 512GB. Tight, only 13GB spare. Est. 38.4 tok/s.
No price yet: Ships late October 2026. Apple has not published a price.
Collecting price history — 3 of 90 days. No timing verdict yet.
A trillion parameters. No Config we track holds it at any quant, and we list it so the arithmetic says so plainly.
Offers — Mac Studio, M5 Ultra, 512GB, 1TB
Ships late October 2026. Apple has not published a price.
- Mac mini, M6, 16GB$1,099Won't fit−478GB headroom
- Mac mini, M6, 24GB$1,299Won't fit−470GB headroom
- Mac mini, M6, 32GB$1,499Won't fit−462GB headroom
- Mac mini, M5 Pro, 24GB$1,699Won't fit−470GB headroom
- Mac mini, M5 Pro, 48GB$2,499Won't fit−446GB headroom
- Mac Studio, M5 Max, 36GB$2,499Won't fit−458GB headroom
- Mac mini, M5 Pro, 64GB$2,899Won't fit−432GB headroom
- GeForce RTX 5090, 32GB$2,999.99+hostWon't fit−460GB headroom
- Won't fit−432GB headroom
- Mac Studio, M5 Max, 48GB$3,099Won't fit−446GB headroom
- Mac Studio, M5 Max, 64GB$3,499Won't fit−432GB headroom
- Won't fit−370GB headroom
- Won't fit−402GB headroom
- Won't fit−371GB headroom
- Won't fit−370GB headroom
- Won't fit−371GB headroom
- Won't fit−402GB headroom
- Won't fit−243GB headroom
- Won't fit−243GB headroom
- Tight13GB headroomest. 38.4 tok/s
| Config | Memory | Price | $/GB | Fit | Headroom | Est. speed |
|---|---|---|---|---|---|---|
| Mac mini, M6 | 16GB | $1,099 | $69 | Won't fit | −478GB | — |
| Mac mini, M6 | 24GB | $1,299 | $54 | Won't fit | −470GB | — |
| Mac mini, M6 | 32GB | $1,499 | $47 | Won't fit | −462GB | — |
| Mac mini, M5 Pro | 24GB | $1,699 | $71 | Won't fit | −470GB | — |
| Mac mini, M5 Pro | 48GB | $2,499 | $52 | Won't fit | −446GB | — |
| Mac Studio, M5 Max | 36GB | $2,499 | $69 | Won't fit | −458GB | — |
| Mac mini, M5 Pro | 64GB | $2,899 | $45 | Won't fit | −432GB | — |
| GeForce RTX 5090 | 32GB | $2,999.99+host | $94 | Won't fit | −460GB | — |
| Mac Studio, M4 Max, refurb | 64GB | $3,049 | $48 | Won't fit | −432GB | — |
| Mac Studio, M5 Max | 48GB | $3,099 | $65 | Won't fit | −446GB | — |
| Mac Studio, M5 Max | 64GB | $3,499 | $55 | Won't fit | −432GB | — |
| Ascent GX10, GB10 Grace Blackwell | 128GB | $3,999.99 | $31 | Won't fit | −370GB | — |
| Mac Studio, M3 Ultra, refurb | 96GB | $4,329 | $45 | Won't fit | −402GB | — |
| Mac Studio, M4 Max, refurb | 128GB | $4,409 | $34 | Won't fit | −371GB | — |
| DGX Spark, GB10 Grace Blackwell | 128GB | $4,699 | $37 | Won't fit | −370GB | — |
| Mac Studio, M5 Max | 128GB | $5,099 | $40 | Won't fit | −371GB | — |
| Mac Studio, M5 Ultra | 96GB | $5,499 | $57 | Won't fit | −402GB | — |
| Mac Studio, M3 Ultra, refurb | 256GB | $7,729 | $30 | Won't fit | −243GB | — |
| Mac Studio, M5 Ultra | 256GB | $9,499 | $37 | Won't fit | −243GB | — |
| Mac Studio, M5 Ultra | 512GB | — | — | Tight | 13GB | 38.4 tok/s |
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 × 0.489 B/param (3.91 bits) + 0.070 GB/1K × 8K K/V + 1.2 GB runtime
speed = bandwidth × 0.5 ÷ (32B active × 0.489 B/param) = 15.6 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.
What else Mac Studio M5 Ultra 512GB runs
- Kimi-K2-InstructQ3_K_M, est. 38.4 tok/s
- DeepSeek-V3Q5_K_M, est. 22.9 tok/s
- DeepSeek-R1Q5_K_M, est. 22.9 tok/s
- Qwen3-Coder-480B-A35B-InstructQ6_K, est. 20.9 tok/s
- Llama-4-Maverick-17B-128EQ8_0, est. 33.2 tok/s
- GLM-4.6Q8_0, est. 17.6 tok/s
Best quant that still fits at 8K context, largest model first. Full list on the Config page.
Bandwidth 1.2TB/s, usable memory 504GB of 512GB.