What GPU do you need to run Laguna XS 2.1?
33.44B params 256K max context Tool useReasoning
The smallest GPU memory that runs Laguna XS 2.1 well is 20 GB (Q4_K_M, e.g. RTX 4000 Ada). 47 of 132 GPUs run it well. 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 70.9
VRAM needed per quantization
| Quantization | Weights | 16K context | 32K context | Full 256K |
|---|---|---|---|---|
| Q4_K_M | 19.1 GB | 19.7 GB | 20.4 GB | 28.9 GB |
| Q8_0 | 33.1 GB | 33.8 GB | 34.4 GB | 42.9 GB |
Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.
Cheapest GPUs that can run Laguna XS 2.1
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 Laguna XS 2.1
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 4000 Ada 20 GB Q4_K_M 22K context · ~132 tok/s
- RTX 3090 24 GB Q4_K_M 127K context · ~344 tok/s
- RTX 3090 Ti 24 GB Q4_K_M 127K context · ~371 tok/s
- RTX 4090 24 GB Q4_K_M 127K context · ~371 tok/s
- RTX 4500 Ada 24 GB Q4_K_M 127K context · ~159 tok/s
- RTX A5000 24 GB Q4_K_M 127K context · ~282 tok/s
- RTX PRO 4000 Blackwell 24 GB Q4_K_M 127K context · ~247 tok/s
- RTX Spark 24 GB Q4_K_M 127K context · ~110 tok/s
- Tesla P40 24 GB Q4_K_M 127K context · ~127 tok/s
- RTX 5090 32 GB Q4_K_M 256K context · ~659 tok/s
- RTX 5000 Ada 32 GB Q4_K_M 256K context · ~212 tok/s
- RTX PRO 4500 Blackwell 32 GB Q4_K_M 256K context · ~329 tok/s
- Tesla V100 32GB 32 GB Q4_K_M 256K context · ~331 tok/s
- L40S 48 GB Q8_0 256K context · ~218 tok/s
- RTX 6000 Ada 48 GB Q8_0 256K context · ~242 tok/s
- RTX A6000 48 GB Q8_0 256K context · ~193 tok/s
- RTX PRO 5000 Blackwell 48 GB Q8_0 256K context · ~339 tok/s
- RTX PRO 5000 Blackwell 72GB 72 GB Q8_0 256K context · ~339 tok/s
- RTX PRO 5500 Blackwell 84 GB Q8_0 256K context · ~352 tok/s
- RTX PRO 6000 Blackwell 96 GB Q8_0 256K context · ~451 tok/s
- DGX Spark (GB 10) 128 GB Q8_0 256K context · ~69 tok/s
AMD
- Radeon RX 7900 XT 20 GB Q4_K_M 22K context · ~294 tok/s
- Radeon RX 7900 XTX 24 GB Q4_K_M 127K context · ~353 tok/s
- Radeon AI PRO R9700 32 GB Q4_K_M 256K context · ~235 tok/s
- Ryzen AI Max (Strix Halo) 32 GB Q4_K_M 256K context · ~94 tok/s
- Ryzen AI Max 400 (Gorgon Halo) 64 GB Q8_0 256K context · ~69 tok/s
Intel
- Arc Pro B65 32 GB Q4_K_M 256K context · ~224 tok/s
- Arc Pro B70 32 GB Q4_K_M 256K context · ~224 tok/s
Apple
- M2 24 GB Q4_K_M 127K context · ~37 tok/s
- M3 24 GB Q4_K_M 127K context · ~37 tok/s
- M4 24 GB Q4_K_M 127K context · ~44 tok/s
- M4 Pro 24 GB Q4_K_M 127K context · ~100 tok/s
- M5 24 GB Q4_K_M 127K context · ~56 tok/s
- M5 Pro 24 GB Q4_K_M 127K context · ~113 tok/s
- M6 24 GB Q4_K_M 127K context · ~63 tok/s
- M1 Max 32 GB Q4_K_M 256K context · ~147 tok/s
- M1 Pro 32 GB Q4_K_M 256K context · ~74 tok/s
- M2 Max 32 GB Q4_K_M 256K context · ~147 tok/s
- M2 Pro 32 GB Q4_K_M 256K context · ~74 tok/s
- M3 Max 36 GB Q8_0 75K context · ~76 tok/s
- M3 Pro 36 GB Q8_0 75K context · ~38 tok/s
- M4 Max 36 GB Q8_0 75K context · ~103 tok/s
- M5 Max 36 GB Q8_0 75K context · ~116 tok/s
- M1 Ultra 64 GB Q8_0 256K context · ~202 tok/s
- M2 Ultra 64 GB Q8_0 256K context · ~202 tok/s
- M3 Ultra 96 GB Q8_0 256K context · ~206 tok/s
- M5 Ultra 96 GB Q8_0 256K context · ~302 tok/s