What GPU do you need to run Flint 40B A2.5B?
39.79B params 256K max context Tool useReasoning
The smallest GPU memory that runs Flint 40B A2.5B well is 32 GB (Q4_K_M, e.g. RTX Spark). 35 of 132 GPUs run it well and 10 more are a tight fit. It's a mixture-of-experts model, so all weights must fit in memory but speed depends only on the active experts.
SWE-bench Verified 74.2
VRAM needed per quantization
| Quantization | Weights | 16K context | 32K context | Full 256K |
|---|---|---|---|---|
| Q4_K_M | 23.6 GB | 24.0 GB | 24.5 GB | 30.5 GB |
| Q8_0 | 39.4 GB | 39.9 GB | 40.3 GB | 46.4 GB |
Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.
Cheapest GPUs that can run Flint 40B A2.5B
Lowest launch MSRP among GPUs that run its smallest quantization (Q4_K_M) well, with at least 16K context.
As an Amazon Associate, I earn from qualifying purchases.
GPUs that can run Flint 40B A2.5B
Best quantization that fits on each GPU, with the context and speed it gets. For chips with several memory sizes, the smallest that works is shown.
NVIDIA
- RTX 3090 24 GB Q4_K_M 14K context · ~377 tok/s tight fit
- RTX 3090 Ti 24 GB Q4_K_M 14K context · ~406 tok/s tight fit
- RTX 4090 24 GB Q4_K_M 14K context · ~406 tok/s tight fit
- RTX 4500 Ada 24 GB Q4_K_M 14K context · ~174 tok/s tight fit
- RTX A5000 24 GB Q4_K_M 14K context · ~310 tok/s tight fit
- RTX PRO 4000 Blackwell 24 GB Q4_K_M 14K context · ~271 tok/s tight fit
- Tesla P40 24 GB Q4_K_M 14K context · ~140 tok/s tight fit
- RTX 5090 32 GB Q4_K_M 256K context · ~723 tok/s
- RTX 5000 Ada 32 GB Q4_K_M 256K context · ~232 tok/s
- RTX PRO 4500 Blackwell 32 GB Q4_K_M 256K context · ~361 tok/s
- RTX Spark 32 GB Q4_K_M 256K context · ~121 tok/s
- Tesla V100 32GB 32 GB Q4_K_M 256K context · ~363 tok/s
- L40S 48 GB Q8_0 256K context · ~248 tok/s
- RTX 6000 Ada 48 GB Q8_0 256K context · ~276 tok/s
- RTX A6000 48 GB Q8_0 256K context · ~221 tok/s
- RTX PRO 5000 Blackwell 48 GB Q8_0 256K context · ~386 tok/s
- RTX PRO 5000 Blackwell 72GB 72 GB Q8_0 256K context · ~386 tok/s
- RTX PRO 5500 Blackwell 84 GB Q8_0 256K context · ~402 tok/s
- RTX PRO 6000 Blackwell 96 GB Q8_0 256K context · ~515 tok/s
- DGX Spark (GB 10) 128 GB Q8_0 256K context · ~78 tok/s
AMD
- Radeon RX 7900 XTX 24 GB Q4_K_M 14K context · ~387 tok/s tight fit
- Radeon AI PRO R9700 32 GB Q4_K_M 256K context · ~258 tok/s
- Ryzen AI Max (Strix Halo) 32 GB Q4_K_M 256K context · ~103 tok/s
- Ryzen AI Max 400 (Gorgon Halo) 64 GB Q8_0 256K context · ~78 tok/s
Intel
- Arc Pro B65 32 GB Q4_K_M 256K context · ~245 tok/s
- Arc Pro B70 32 GB Q4_K_M 256K context · ~245 tok/s
Apple
- M2 24 GB Q4_K_M 14K context · ~40 tok/s tight fit
- M3 24 GB Q4_K_M 14K context · ~40 tok/s tight fit
- M1 Max 32 GB Q4_K_M 256K context · ~161 tok/s
- M1 Pro 32 GB Q4_K_M 256K context · ~81 tok/s
- M2 Max 32 GB Q4_K_M 256K context · ~161 tok/s
- M2 Pro 32 GB Q4_K_M 256K context · ~81 tok/s
- M4 32 GB Q4_K_M 256K context · ~48 tok/s
- M5 32 GB Q4_K_M 256K context · ~62 tok/s
- M6 32 GB Q4_K_M 256K context · ~69 tok/s
- M3 Max 36 GB Q4_K_M 256K context · ~121 tok/s
- M3 Pro 36 GB Q4_K_M 256K context · ~60 tok/s
- M4 Max 36 GB Q4_K_M 256K context · ~165 tok/s
- M5 Max 36 GB Q4_K_M 256K context · ~185 tok/s
- M4 Pro 48 GB Q8_0 256K context · ~78 tok/s
- M5 Pro 48 GB Q8_0 256K context · ~88 tok/s
- M1 Ultra 64 GB Q8_0 256K context · ~230 tok/s
- M2 Ultra 64 GB Q8_0 256K context · ~230 tok/s
- M3 Ultra 96 GB Q8_0 256K context · ~235 tok/s
- M5 Ultra 96 GB Q8_0 256K context · ~345 tok/s