Poolside Laguna S 2.1
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q4_K_M poolside/Laguna-S-2.1-GGUF | 89.44 GiB | 1 | 90.9 GiB+ | 96 GB GPU | HF files |
| Q8_0 poolside/Laguna-S-2.1-GGUF | 120 GiB | 1 | 121 GiB+ | 128 GB GPU | HF files |
| FP16 poolside/Laguna-S-2.1-GGUF | 219 GiB | 1 | 220 GiB+ | over 192 GB GPU | HF files |
| BF16 unsloth/Laguna-S-2.1-GGUF | 219 GiB | 5 | 220 GiB+ | over 192 GB GPU | HF files |
| MLX 4-bit mixed precision TensorFold/Laguna-S-2.1-MLX-4bit | 61.60 GiB | 13 | 63.1 GiB+ | 96 GB Mac | HF files |
| MLX 8-bit mixed precision TensorFold/Laguna-S-2.1-MLX-8bit | 116 GiB | 24 | 118 GiB+ | 192 GB Mac | HF files |
No verified file: Q5_K_M. Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory.
Quality
Artificial Analysis Intelligence Index
Unknown
Arena text rating
Unknown
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 9
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| NVIDIA DGX Spark (GB10, 128 GB) | Q4_K_M | Ollama (llama.cpp CUDA) | — | 12.6 | 2026-07-21 | huggingface.co |
| Apple M5 Max (40-core GPU), 128 GB unified memory | MLX oQ4e (oMLX custom ~4-bit) | MLX | 1,517 | 58.2 | 2026-10-09 | omlx.ai |
| Apple M3 Max (40-core GPU), 128 GB unified memory | MLX oQ4e (oMLX custom ~4-bit) | MLX | 406 | 39 | 2026-10-02 | omlx.ai |
| Apple M4 Max (40-core GPU), 128 GB unified memory | MLX oQ4e (oMLX custom ~4-bit) | MLX | 506 | 52 | 2026-08-25 | omlx.ai |
| Apple M3 Ultra (60-core GPU), 256 GB unified memory | MLX oQ4e (oMLX custom ~4-bit) | MLX | 631 | 53.9 | 2026-08-22 | omlx.ai |
| Apple M3 Ultra (60-core GPU), 256 GB unified memory | MLX oQ4e (oMLX custom ~4-bit) | MLX | 535 | 45.8 | 2026-08-22 | omlx.ai |
| Apple M2 Ultra (76-core GPU), 192 GB unified memory | MLX oQ4e (oMLX custom ~4-bit) | MLX | 402 | 49.4 | 2026-07-25 | omlx.ai |
| Apple M1 Ultra (48-core GPU), 128 GB unified memory | MLX oQ4e (oMLX custom ~4-bit) | MLX | 245 | 32.8 | 2026-07-22 | omlx.ai |
| Apple Silicon Mac Studio (chip not stated, 128 GB class) | Q4_K_M | Ollama (llama.cpp Metal) | — | 17.6 | 2026-07-21 | huggingface.co |
Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database