Cheapest box to run Gemma-4-26B-A4B at Q3_K_M
Google’s Gemma-4-26B-A4B is 25.2B parameters, 3.8B 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.
Gemma-4-26B-A4B at Q3_K_M needs 14GB.
Cheapest Config that fits: Mac mini, M6, 24GB. Fits, 7GB headroom. Est. 45.8 tok/s.
Best offer today: $1,299 at Best Buy — in stock, checked 33 days ago.
Price history: 34 days recorded.
The 26B in the name is marketing; Google's own table reads 25.2B total and 3.8B active, and that is what we publish. Only the 5 global-attention layers cache unboundedly; the rest slide a 1024-token window.
Offers — Mac mini, M6, 24GB, 512GB
- Mac mini, M6, 16GB$1,099Won't fit−1GB headroom
- Mac mini, M6, 24GB$1,299Fits7GB headroomest. 45.8 tok/s
- Mac mini, M6, 32GB$1,499Fits15GB headroomest. 45.8 tok/s
- Mac mini, M5 Pro, 24GB$1,699Fits7GB headroomest. 82.6 tok/s
- Mac mini, M5 Pro, 48GB$2,499Fits30GB headroomest. 82.6 tok/s
- Mac Studio, M5 Max, 36GB$2,499Fits19GB headroomest. 124 tok/s
- Mac mini, M5 Pro, 64GB$2,899Fits45GB headroomest. 82.6 tok/s
- GeForce RTX 5090, 32GB$2,999.99+hostFits17GB headroomest. 482 tok/s
- Fits45GB headroomest. 147 tok/s
- Mac Studio, M5 Max, 48GB$3,099Fits30GB headroomest. 165 tok/s
- Mac Studio, M5 Max, 64GB$3,499Fits45GB headroomest. 165 tok/s
- Fits106GB headroomest. 73.5 tok/s
- Fits74GB headroomest. 220 tok/s
- Fits106GB headroomest. 147 tok/s
- Fits106GB headroomest. 73.5 tok/s
- Fits106GB headroomest. 165 tok/s
- Fits74GB headroomest. 323 tok/s
- Fits234GB headroomest. 220 tok/s
- Fits234GB headroomest. 323 tok/s
- Fits490GB headroomest. 323 tok/s
| Config | Memory | Price | $/GB | Fit | Headroom | Est. speed |
|---|---|---|---|---|---|---|
| Mac mini, M6 | 16GB | $1,099 | $69 | Won't fit | −1GB | — |
| Mac mini, M6 | 24GB | $1,299 | $54 | Fits | 7GB | 45.8 tok/s |
| Mac mini, M6 | 32GB | $1,499 | $47 | Fits | 15GB | 45.8 tok/s |
| Mac mini, M5 Pro | 24GB | $1,699 | $71 | Fits | 7GB | 82.6 tok/s |
| Mac mini, M5 Pro | 48GB | $2,499 | $52 | Fits | 30GB | 82.6 tok/s |
| Mac Studio, M5 Max | 36GB | $2,499 | $69 | Fits | 19GB | 124 tok/s |
| Mac mini, M5 Pro | 64GB | $2,899 | $45 | Fits | 45GB | 82.6 tok/s |
| GeForce RTX 5090 | 32GB | $2,999.99+host | $94 | Fits | 17GB | 482 tok/s |
| Mac Studio, M4 Max, refurb | 64GB | $3,049 | $48 | Fits | 45GB | 147 tok/s |
| Mac Studio, M5 Max | 48GB | $3,099 | $65 | Fits | 30GB | 165 tok/s |
| Mac Studio, M5 Max | 64GB | $3,499 | $55 | Fits | 45GB | 165 tok/s |
| Ascent GX10, GB10 Grace Blackwell | 128GB | $3,999.99 | $31 | Fits | 106GB | 73.5 tok/s |
| Mac Studio, M3 Ultra, refurb | 96GB | $4,329 | $45 | Fits | 74GB | 220 tok/s |
| Mac Studio, M4 Max, refurb | 128GB | $4,409 | $34 | Fits | 106GB | 147 tok/s |
| DGX Spark, GB10 Grace Blackwell | 128GB | $4,699 | $37 | Fits | 106GB | 73.5 tok/s |
| Mac Studio, M5 Max | 128GB | $5,099 | $40 | Fits | 106GB | 165 tok/s |
| Mac Studio, M5 Ultra | 96GB | $5,499 | $57 | Fits | 74GB | 323 tok/s |
| Mac Studio, M3 Ultra, refurb | 256GB | $7,729 | $30 | Fits | 234GB | 220 tok/s |
| Mac Studio, M5 Ultra | 256GB | $9,499 | $37 | Fits | 234GB | 323 tok/s |
| Mac Studio, M5 Ultra | 512GB | — | — | Fits | 490GB | 323 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 = 25.2B × 0.489 B/param (3.91 bits) + 0.042 GB/1K × 8K K/V + 1.2 GB runtime
speed = bandwidth × 0.5 ÷ (3.8B active × 0.489 B/param) = 1.9 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 mini M6 24GB runs
- Qwen3.5-35B-A3BQ3_K_M, est. 58.0 tok/s
- Qwen3-32BQ3_K_M, est. 7.1 tok/s
- Gemma-4-31BQ3_K_M, est. 7.4 tok/s
- Qwen3-30B-A3BQ4_K_M, est. 42.7 tok/s
- Gemma-3-27BQ4_K_M, est. 6.8 tok/s
- Qwen3.6-27BQ5_K_M, est. 5.8 tok/s
Best quant that still fits at 8K context, largest model first. Full list on the Config page.
Bandwidth 170GB/s, usable memory 21GB of 24GB.