Cheapest box to run Qwen3-30B-A3B at BF16
Alibaba’s Qwen3-30B-A3B is 30.5B parameters, 3.3B 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.
Qwen3-30B-A3B at BF16 needs 63GB.
Cheapest Config that fits: Ascent GX10, GB10 Grace Blackwell, 128GB. Fits, 57GB headroom. Est. 20.7 tok/s.
Best offer today: $3,999.99 at Best Buy — in stock, checked 37 hours ago.
Collecting price history — 3 of 90 days. No timing verdict yet.
Offers — Ascent GX10, GB10 Grace Blackwell, 128GB, 1TB
- Mac mini, M6, 16GB$1,099Won't fit−50GB headroom
- Mac mini, M6, 24GB$1,299Won't fit−42GB headroom
- Mac mini, M6, 32GB$1,499Won't fit−34GB headroom
- Mac mini, M5 Pro, 24GB$1,699Won't fit−42GB headroom
- Mac mini, M5 Pro, 48GB$2,499Won't fit−19GB headroom
- Mac Studio, M5 Max, 36GB$2,499Won't fit−30GB headroom
- Mac mini, M5 Pro, 64GB$2,899Won't fit−4GB headroom
- GeForce RTX 5090, 32GB$2,999.99+hostWon't fit−32GB headroom
- Won't fit−4GB headroom
- Mac Studio, M5 Max, 48GB$3,099Won't fit−19GB headroom
- Mac Studio, M5 Max, 64GB$3,499Won't fit−4GB headroom
- Fits57GB headroomest. 20.7 tok/s
- Fits25GB headroomest. 62.0 tok/s
- Fits57GB headroomest. 41.4 tok/s
- Fits57GB headroomest. 20.7 tok/s
- Fits57GB headroomest. 46.5 tok/s
- Fits25GB headroomest. 90.9 tok/s
- Fits185GB headroomest. 62.0 tok/s
- Fits185GB headroomest. 90.9 tok/s
- Fits441GB headroomest. 90.9 tok/s
| Config | Memory | Price | $/GB | Fit | Headroom | Est. speed |
|---|---|---|---|---|---|---|
| Mac mini, M6 | 16GB | $1,099 | $69 | Won't fit | −50GB | — |
| Mac mini, M6 | 24GB | $1,299 | $54 | Won't fit | −42GB | — |
| Mac mini, M6 | 32GB | $1,499 | $47 | Won't fit | −34GB | — |
| Mac mini, M5 Pro | 24GB | $1,699 | $71 | Won't fit | −42GB | — |
| Mac mini, M5 Pro | 48GB | $2,499 | $52 | Won't fit | −19GB | — |
| Mac Studio, M5 Max | 36GB | $2,499 | $69 | Won't fit | −30GB | — |
| Mac mini, M5 Pro | 64GB | $2,899 | $45 | Won't fit | −4GB | — |
| GeForce RTX 5090 | 32GB | $2,999.99+host | $94 | Won't fit | −32GB | — |
| Mac Studio, M4 Max, refurb | 64GB | $3,049 | $48 | Won't fit | −4GB | — |
| Mac Studio, M5 Max | 48GB | $3,099 | $65 | Won't fit | −19GB | — |
| Mac Studio, M5 Max | 64GB | $3,499 | $55 | Won't fit | −4GB | — |
| Ascent GX10, GB10 Grace Blackwell | 128GB | $3,999.99 | $31 | Fits | 57GB | 20.7 tok/s |
| Mac Studio, M3 Ultra, refurb | 96GB | $4,329 | $45 | Fits | 25GB | 62.0 tok/s |
| Mac Studio, M4 Max, refurb | 128GB | $4,409 | $34 | Fits | 57GB | 41.4 tok/s |
| DGX Spark, GB10 Grace Blackwell | 128GB | $4,699 | $37 | Fits | 57GB | 20.7 tok/s |
| Mac Studio, M5 Max | 128GB | $5,099 | $40 | Fits | 57GB | 46.5 tok/s |
| Mac Studio, M5 Ultra | 96GB | $5,499 | $57 | Fits | 25GB | 90.9 tok/s |
| Mac Studio, M3 Ultra, refurb | 256GB | $7,729 | $30 | Fits | 185GB | 62.0 tok/s |
| Mac Studio, M5 Ultra | 256GB | $9,499 | $37 | Fits | 185GB | 90.9 tok/s |
| Mac Studio, M5 Ultra | 512GB | — | — | Fits | 441GB | 90.9 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 = 30.5B × 2.000 B/param (16 bits) + 0.098 GB/1K × 8K K/V + 1.2 GB runtime
speed = bandwidth × 0.5 ÷ (3.3B active × 2.000 B/param) = 6.6 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 Ascent GX10 GB10 Grace Blackwell 128GB runs
- Qwen3-235B-A22BQ3_K_M, est. 12.7 tok/s
- MiniMax-M2Q3_K_M, est. 27.9 tok/s
- Mistral-Large-2411Q6_K, est. 1.8 tok/s
- gpt-oss-120bQ6_K, est. 32.6 tok/s
- Llama-4-Scout-17B-16EQ8_0, est. 7.6 tok/s
- Llama-3.3-70BQ8_0, est. 2.4 tok/s
Best quant that still fits at 8K context, largest model first. Full list on the Config page.
Bandwidth 273GB/s, usable memory 120GB of 128GB.