LFM3-VL 2B VRAM requirements: which GPUs can run it locally?
The smallest GPU memory that runs LFM3-VL 2B well is 2 GB (Q8_0, e.g. T400). 132 of 132 GPUs run it well.
VRAM needed per quantization
| Quantization | Weights | 16K context | 32K context | Full 128K |
|---|---|---|---|---|
| Q4_K_M | 1.0 GB | 1.2 GB | 1.4 GB | 2.4 GB |
| Q8_0 | 1.7 GB | 1.9 GB | 2.1 GB | 3.1 GB |
Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.
Run it locally
Each command downloads the GGUF from Hugging Face, then Ollama starts a chat and llama.cpp starts an OpenAI-compatible server with a web UI on port 8080. Both fetch the vision projector (mmproj) automatically.
ollama run hf.co/LiquidAI/LFM3-VL-2B-GGUF:Q4_K_M llama-server -hf LiquidAI/LFM3-VL-2B-GGUF:Q4_K_M ollama run hf.co/LiquidAI/LFM3-VL-2B-GGUF:Q8_0 llama-server -hf LiquidAI/LFM3-VL-2B-GGUF:Q8_0 Cheapest GPUs that can run LFM3-VL 2B
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 LFM3-VL 2B
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
- T400 2 GB Q8_0 25K context · ~28 tok/s
- T1000 4 GB Q8_0 128K context · ~56 tok/s
- T400 4GB 4 GB Q8_0 128K context · ~28 tok/s
- T600 4 GB Q8_0 128K context · ~56 tok/s
- GTX 1060 6GB 6 GB Q8_0 128K context · ~67 tok/s
- GTX 1660 6 GB Q8_0 128K context · ~67 tok/s
- GTX 1660 SUPER 6 GB Q8_0 128K context · ~118 tok/s
- GTX 1660 Ti 6 GB Q8_0 128K context · ~101 tok/s
- RTX 2060 6 GB Q8_0 128K context · ~118 tok/s
- RTX 3050 6GB 6 GB Q8_0 128K context · ~59 tok/s
- RTX A2000 6 GB Q8_0 128K context · ~101 tok/s
- GTX 1070 8 GB Q8_0 128K context · ~90 tok/s
- GTX 1070 Ti 8 GB Q8_0 128K context · ~90 tok/s
- GTX 1080 8 GB Q8_0 128K context · ~112 tok/s
- RTX 2060 SUPER 8 GB Q8_0 128K context · ~157 tok/s
- RTX 2070 8 GB Q8_0 128K context · ~157 tok/s
- RTX 2070 SUPER 8 GB Q8_0 128K context · ~157 tok/s
- RTX 2080 8 GB Q8_0 128K context · ~157 tok/s
- RTX 2080 SUPER 8 GB Q8_0 128K context · ~174 tok/s
- RTX 3050 8 GB Q8_0 128K context · ~79 tok/s
- RTX 3060 Ti 8 GB Q8_0 128K context · ~157 tok/s
- RTX 3060 Ti GDDR6X 8 GB Q8_0 128K context · ~214 tok/s
- RTX 3070 8 GB Q8_0 128K context · ~157 tok/s
- RTX 3070 Ti 8 GB Q8_0 128K context · ~214 tok/s
- RTX 4060 8 GB Q8_0 128K context · ~96 tok/s
- RTX 4060 Ti 8 GB Q8_0 128K context · ~101 tok/s
- RTX 5050 8 GB Q8_0 128K context · ~112 tok/s
- RTX 5060 8 GB Q8_0 128K context · ~157 tok/s
- RTX 5060 Ti 8 GB Q8_0 128K context · ~157 tok/s
- T1000 8GB 8 GB Q8_0 128K context · ~56 tok/s
- RTX 3080 10 GB Q8_0 128K context · ~267 tok/s
- GTX 1080 Ti 11 GB Q8_0 128K context · ~170 tok/s
- RTX 2080 Ti 11 GB Q8_0 128K context · ~216 tok/s
- RTX 2060 12GB 12 GB Q8_0 128K context · ~118 tok/s
- RTX 3060 12 GB Q8_0 128K context · ~126 tok/s
- RTX 3080 12GB 12 GB Q8_0 128K context · ~320 tok/s
- RTX 3080 Ti 12 GB Q8_0 128K context · ~320 tok/s
- RTX 4070 12 GB Q8_0 128K context · ~177 tok/s
- RTX 4070 SUPER 12 GB Q8_0 128K context · ~177 tok/s
- RTX 4070 Ti 12 GB Q8_0 128K context · ~177 tok/s
- RTX 5070 12 GB Q8_0 128K context · ~236 tok/s
- RTX A2000 12GB 12 GB Q8_0 128K context · ~101 tok/s
- RTX 4060 Ti 16GB 16 GB Q8_0 128K context · ~101 tok/s
- RTX 4070 Ti SUPER 16 GB Q8_0 128K context · ~236 tok/s
- RTX 4080 16 GB Q8_0 128K context · ~252 tok/s
- RTX 4080 SUPER 16 GB Q8_0 128K context · ~259 tok/s
- RTX 5060 Ti 16GB 16 GB Q8_0 128K context · ~157 tok/s
- RTX 5070 Ti 16 GB Q8_0 128K context · ~315 tok/s
- RTX 5080 16 GB Q8_0 128K context · ~337 tok/s
- RTX 2000 Ada 16 GB Q8_0 128K context · ~79 tok/s
- RTX A4000 16 GB Q8_0 128K context · ~157 tok/s
- RTX PRO 2000 Blackwell 16 GB Q8_0 128K context · ~101 tok/s
- Tesla T4 16 GB Q8_0 128K context · ~112 tok/s
- Tesla V100 16GB 16 GB Q8_0 128K context · ~316 tok/s
- RTX 4000 Ada 20 GB Q8_0 128K context · ~126 tok/s
- RTX 3090 24 GB Q8_0 128K context · ~329 tok/s
- RTX 3090 Ti 24 GB Q8_0 128K context · ~354 tok/s
- RTX 4090 24 GB Q8_0 128K context · ~354 tok/s
- RTX 4500 Ada 24 GB Q8_0 128K context · ~152 tok/s
- RTX A5000 24 GB Q8_0 128K context · ~270 tok/s
- RTX PRO 4000 Blackwell 24 GB Q8_0 128K context · ~236 tok/s
- RTX Spark 24 GB Q8_0 128K context · ~105 tok/s
- Tesla P40 24 GB Q8_0 128K context · ~122 tok/s
- RTX 5090 32 GB Q8_0 128K context · ~629 tok/s
- RTX 5000 Ada 32 GB Q8_0 128K context · ~202 tok/s
- RTX PRO 4500 Blackwell 32 GB Q8_0 128K context · ~315 tok/s
- Tesla V100 32GB 32 GB Q8_0 128K context · ~316 tok/s
- L40S 48 GB Q8_0 128K context · ~303 tok/s
- RTX 6000 Ada 48 GB Q8_0 128K context · ~337 tok/s
- RTX A6000 48 GB Q8_0 128K context · ~270 tok/s
- RTX PRO 5000 Blackwell 48 GB Q8_0 128K context · ~472 tok/s
- RTX PRO 5000 Blackwell 72GB 72 GB Q8_0 128K context · ~472 tok/s
- RTX PRO 5500 Blackwell 84 GB Q8_0 128K context · ~491 tok/s
- RTX PRO 6000 Blackwell 96 GB Q8_0 128K context · ~629 tok/s
- DGX Spark (GB 10) 128 GB Q8_0 128K context · ~96 tok/s
AMD
- Radeon RX 5600 XT 6 GB Q8_0 128K context · ~101 tok/s
- Radeon RX 5700 8 GB Q8_0 128K context · ~157 tok/s
- Radeon RX 5700 XT 8 GB Q8_0 128K context · ~157 tok/s
- Radeon RX 6600 8 GB Q8_0 128K context · ~79 tok/s
- Radeon RX 6600 XT 8 GB Q8_0 128K context · ~90 tok/s
- Radeon RX 6650 XT 8 GB Q8_0 128K context · ~98 tok/s
- Radeon RX 7600 8 GB Q8_0 128K context · ~101 tok/s
- Radeon RX 9060 8 GB Q8_0 128K context · ~101 tok/s
- Radeon RX 9060 XT 8 GB Q8_0 128K context · ~112 tok/s
- Radeon RX 6700 XT 12 GB Q8_0 128K context · ~135 tok/s
- Radeon RX 6750 XT 12 GB Q8_0 128K context · ~152 tok/s
- Radeon RX 7700 XT 12 GB Q8_0 128K context · ~152 tok/s
- Radeon RX 9070 GRE 12 GB Q8_0 128K context · ~152 tok/s
- AMD BC-250 16 GB Q8_0 128K context · ~157 tok/s
- Radeon RX 6800 16 GB Q8_0 128K context · ~180 tok/s
- Radeon RX 6800 XT 16 GB Q8_0 128K context · ~180 tok/s
- Radeon RX 6900 XT 16 GB Q8_0 128K context · ~180 tok/s
- Radeon RX 6950 XT 16 GB Q8_0 128K context · ~202 tok/s
- Radeon RX 7600 XT 16 GB Q8_0 128K context · ~101 tok/s
- Radeon RX 7800 XT 16 GB Q8_0 128K context · ~219 tok/s
- Radeon RX 7900 GRE 16 GB Q8_0 128K context · ~202 tok/s
- Radeon RX 9060 XT 16GB 16 GB Q8_0 128K context · ~112 tok/s
- Radeon RX 9070 16 GB Q8_0 128K context · ~225 tok/s
- Radeon RX 9070 XT 16 GB Q8_0 128K context · ~225 tok/s
- Radeon RX 7900 XT 20 GB Q8_0 128K context · ~281 tok/s
- Radeon RX 7900 XTX 24 GB Q8_0 128K context · ~337 tok/s
- Radeon AI PRO R9700 32 GB Q8_0 128K context · ~225 tok/s
- Ryzen AI Max (Strix Halo) 32 GB Q8_0 128K context · ~90 tok/s
- Ryzen AI Max 400 (Gorgon Halo) 64 GB Q8_0 128K context · ~96 tok/s
Intel
- Arc A580 8 GB Q8_0 128K context · ~180 tok/s
- Arc A750 8 GB Q8_0 128K context · ~180 tok/s
- Arc A770 8GB 8 GB Q8_0 128K context · ~180 tok/s
- Arc B570 10 GB Q8_0 128K context · ~133 tok/s
- Arc B580 12 GB Q8_0 128K context · ~160 tok/s
- Arc A770 16GB 16 GB Q8_0 128K context · ~197 tok/s
- Arc Pro B65 32 GB Q8_0 128K context · ~214 tok/s
- Arc Pro B70 32 GB Q8_0 128K context · ~214 tok/s
Apple
- M1 8 GB Q8_0 128K context · ~24 tok/s
- M2 8 GB Q8_0 128K context · ~35 tok/s
- M3 8 GB Q8_0 128K context · ~35 tok/s
- M1 Pro 16 GB Q8_0 128K context · ~70 tok/s
- M2 Pro 16 GB Q8_0 128K context · ~70 tok/s
- M4 16 GB Q8_0 128K context · ~42 tok/s
- M5 16 GB Q8_0 128K context · ~54 tok/s
- M6 16 GB Q8_0 128K context · ~54 tok/s
- M3 Pro 18 GB Q8_0 128K context · ~53 tok/s
- M4 Pro 24 GB Q8_0 128K context · ~96 tok/s
- M5 Pro 24 GB Q8_0 128K context · ~108 tok/s
- M1 Max 32 GB Q8_0 128K context · ~141 tok/s
- M2 Max 32 GB Q8_0 128K context · ~141 tok/s
- M3 Max 36 GB Q8_0 128K context · ~105 tok/s
- M4 Max 36 GB Q8_0 128K context · ~144 tok/s
- M5 Max 36 GB Q8_0 128K context · ~162 tok/s
- M1 Ultra 64 GB Q8_0 128K context · ~197 tok/s
- M2 Ultra 64 GB Q8_0 128K context · ~197 tok/s
- M3 Ultra 96 GB Q8_0 128K context · ~201 tok/s
- M5 Ultra 96 GB Q8_0 128K context · ~295 tok/s