unsloth
LFM2.5 VL 3B
LFM2.5 VL 3B is a 2.70B dense lfm2 model from unsloth. Configured context length is 128,000 tokens. License: other. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters2.7B
Context125K
Attentiongqa
Licenseother
30 layers · 8 KV heads · head dim 64 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 5.03 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 3.21 GB | Hugging Face file bytes |
| Q8_0 | 2.68 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 2.24 GB | Hugging Face file bytes |
| Q6_K | 2.07 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 1.81 GB | Hugging Face file bytes |
| Q5_K_M | 1.81 GB | Hugging Face file bytes |
| Q5_K_S | 1.77 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 1.60 GB | Hugging Face file bytes |
| Q4_K_M | 1.56 GB | Hugging Face file bytes |
| Q4_K_S | 1.49 GB | Hugging Face file bytes |
| IQ4_NL | 1.48 GB | Hugging Face file bytes |
| IQ4_XS | 1.41 GB | Hugging Face file bytes |
| Q4_1 | 1.63 GB | Hugging Face file bytes |
| Q4_0 | 1.49 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 1.32 GB | Hugging Face file bytes |
| Q3_K_M | 1.27 GB | Hugging Face file bytes |
| Q3_K_S | 1.18 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 1.06 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 1.05 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Runs | BF16 | 27–37 |
| GeForce RTX 3060 12GB | Runs great | BF16 | 34–46 |
| GeForce RTX 4060 Ti 16GB | Runs great | BF16 | 27–37 |
| GeForce RTX 4090 | Runs great | BF16 | 94–128 |
| GeForce RTX 5090 | Runs great | BF16 | 167–228 |
| Apple M4 Pro | Runs great | BF16 | 25–35 |
| Apple M5 | Runs great | BF16 | 14–19 |
| CPU only · DDR5-5600 | Runs on CPU | BF16 | 8.3–11 |
Smallest devices that fit it fully
- GeForce RTX 3050 6GB — Runs, UD-Q8_K_XL, 23–32 tok/s
- GeForce RTX 2060 6GB — Runs, UD-Q8_K_XL, 47–64 tok/s
- GeForce GTX 1660 Super — Runs, UD-Q8_K_XL, 47–64 tok/s
- GeForce GTX 1660 Ti — Runs, UD-Q8_K_XL, 40–55 tok/s
- Intel Arc A380 — Runs, UD-Q8_K_XL, 26–35 tok/s
- GeForce RTX 5060 Ti 8GB — Runs, BF16, 42–57 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (BF16).
llama.cpp
llama-server -hf unsloth/LFM2.5-VL-3B-GGUF:BF16 -c 8192 -ngl 30
Ollama
ollama run lfm2.5
LM Studio
lms get unsloth/LFM2.5-VL-3B-GGUF # Then load quant BF16