LFM2 1.2B VRAM requirements
| 4-bit weights: 677 MB | With 8K context: 1.5 GB | With 32K context: 2.3 GB | 8-bit, 8K context: 2.0 GB |
- Parameters
- 1.2B
- Longest context
- 125K
- Released
- Jul 2025
- License
- Lfm1.0
- Strongest at
- Not rated
Its licence (Lfm1.0) is not a standard open-source one: check the terms before commercial use.
How much VRAM LFM2 1.2B needs
| Precision | Weights | With 8K | With 32K | With 125K |
|---|---|---|---|---|
| 16-bit | 2.2 GB | 3.1 GB | 3.8 GB | 6.7 GB |
| 8-bit | 1.2 GB | 2.0 GB | 2.8 GB | 5.7 GB |
| 6-bit | 915 MB | 1.8 GB | 2.5 GB | 5.4 GB |
| 5-bit | 794 MB | 1.7 GB | 2.4 GB | 5.3 GB |
| 4-bit | 677 MB | 1.5 GB | 2.3 GB | 5.2 GB |
| 3-bit | 546 MB | 1.4 GB | 2.2 GB | 5.1 GB |
| 2-bit | 419 MB | 1.3 GB | 2.0 GB | 5.0 GB |
| 1.6-bit | 230 MB | 1.1 GB | 1.9 GB | 4.8 GB |
One conversation, 16-bit context cache, runtime overhead included.
GPUs and Macs that run LFM2 1.2B
| Machine | Best fit | Context | Speed | First reply | Report |
|---|---|---|---|---|---|
| RTX 4060 | 8-bit | 125K | 160 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 4060 report |
| RTX 5060 | 8-bit | 125K | 251 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 5060 report |
| RTX 3070 | 8-bit | 125K | 251 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 3070 report |
| RTX 3080 10 GB | 8-bit | 125K | 392 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 3080 10 GB report |
| RTX 3060 12 GB | 8-bit | 125K | 206 tok/s | 0.2 s | PDF: LFM2 1.2B on RTX 3060 12 GB report |
| Arc B580 | 8-bit | 125K | 173 tok/s | 0.1 s | PDF: LFM2 1.2B on Arc B580 report |
| RTX 4070 | 8-bit | 125K | 278 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 4070 report |
| RTX 4070 Super | 8-bit | 125K | 278 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 4070 Super report |
| RTX 5070 | 8-bit | 125K | 355 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 5070 report |
| RTX 4060 Ti 16 GB | 8-bit | 125K | 168 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 4060 Ti 16 GB report |
| RTX 5060 Ti 16 GB | 8-bit | 125K | 251 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 5060 Ti 16 GB report |
| Arc A770 16 GB | 8-bit | 125K | 204 tok/s | 0.1 s | PDF: LFM2 1.2B on Arc A770 16 GB report |
| Radeon RX 9070 XT | 8-bit | 125K | 273 tok/s | 0.1 s | PDF: LFM2 1.2B on Radeon RX 9070 XT report |
| RTX 4070 Ti Super | 8-bit | 125K | 355 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 4070 Ti Super report |
| RTX 4080 Super | 8-bit | 125K | 382 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 4080 Super report |
| RTX 5070 Ti | 8-bit | 125K | 447 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 5070 Ti report |
| RTX 5080 | 8-bit | 125K | 472 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 5080 report |
| Radeon RX 7900 XTX | 8-bit | 125K | 369 tok/s | 0.1 s | PDF: LFM2 1.2B on Radeon RX 7900 XTX report |
| RTX 3090 | 8-bit | 125K | 463 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 3090 report |
| RTX 4090 | 8-bit | 125K | 490 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 4090 report |
| Radeon AI PRO R9700 | 8-bit | 125K | 273 tok/s | 0.1 s | PDF: LFM2 1.2B on Radeon AI PRO R9700 report |
| RTX 5090 | 8-bit | 125K | 737 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 5090 report |
| 2× RTX 3090 | 8-bit | 125K | 463 tok/s | 0.1 s | PDF: LFM2 1.2B on 2× RTX 3090 report |
| 2× RTX 4090 | 8-bit | 125K | 490 tok/s | 0.1 s | PDF: LFM2 1.2B on 2× RTX 4090 report |
| RTX PRO 5000 Blackwell 48 GB | 8-bit | 125K | 606 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX PRO 5000 Blackwell 48 GB report |
| RTX 6000 Ada | 8-bit | 125K | 472 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX 6000 Ada report |
| L40S | 8-bit | 125K | 435 tok/s | 0.1 s | PDF: LFM2 1.2B on L40S report |
| 2× RTX 5090 | 8-bit | 125K | 737 tok/s | 0.1 s | PDF: LFM2 1.2B on 2× RTX 5090 report |
| 4× RTX 3090 | 8-bit | 125K | 463 tok/s | 0.1 s | PDF: LFM2 1.2B on 4× RTX 3090 report |
| RTX PRO 6000 Blackwell | 8-bit | 125K | 737 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX PRO 6000 Blackwell report |
| Mac mini M4 16 GB | 8-bit | 125K | 80 tok/s | 1.3 s | PDF: LFM2 1.2B on Mac mini M4 16 GB report |
| Mac mini M4 24 GB | 8-bit | 125K | 80 tok/s | 1.3 s | PDF: LFM2 1.2B on Mac mini M4 24 GB report |
| Mac mini M4 32 GB | 8-bit | 125K | 80 tok/s | 1.3 s | PDF: LFM2 1.2B on Mac mini M4 32 GB report |
| Mac mini M4 Pro 24 GB | 8-bit | 125K | 161 tok/s | 0.7 s | PDF: LFM2 1.2B on Mac mini M4 Pro 24 GB report |
| Mac mini M4 Pro 48 GB | 8-bit | 125K | 161 tok/s | 0.7 s | PDF: LFM2 1.2B on Mac mini M4 Pro 48 GB report |
| Mac mini M4 Pro 64 GB | 8-bit | 125K | 161 tok/s | 0.7 s | PDF: LFM2 1.2B on Mac mini M4 Pro 64 GB report |
| M4 Max 128 GB | 8-bit | 125K | 252 tok/s | 0.3 s | PDF: LFM2 1.2B on M4 Max 128 GB report |
| M5 Max 128 GB | 8-bit | 125K | 272 tok/s | 0.1 s | PDF: LFM2 1.2B on M5 Max 128 GB report |
| M3 Ultra 96 GB | 8-bit | 125K | 299 tok/s | 0.2 s | PDF: LFM2 1.2B on M3 Ultra 96 GB report |
| M3 Ultra 256 GB | 8-bit | 125K | 299 tok/s | 0.2 s | PDF: LFM2 1.2B on M3 Ultra 256 GB report |
| M3 Ultra 512 GB | 8-bit | 125K | 299 tok/s | 0.2 s | PDF: LFM2 1.2B on M3 Ultra 512 GB report |
| M5 Ultra 512 GB | 8-bit | 125K | 372 tok/s | 0.1 s | PDF: LFM2 1.2B on M5 Ultra 512 GB report |
| Ryzen AI Max+ 395 64 GB | 8-bit | 125K | 126 tok/s | 0.2 s | PDF: LFM2 1.2B on Ryzen AI Max+ 395 64 GB report |
| Ryzen AI Max+ 395 128 GB | 8-bit | 125K | 126 tok/s | 0.2 s | PDF: LFM2 1.2B on Ryzen AI Max+ 395 128 GB report |
| RTX Spark | 8-bit | 125K | 175 tok/s | 0.1 s | PDF: LFM2 1.2B on RTX Spark report |
| DGX Spark | 8-bit | 125K | 160 tok/s | 0.1 s | PDF: LFM2 1.2B on DGX Spark report |
| Dual DGX Spark | 8-bit | 125K | 160 tok/s | 0.1 s | PDF: LFM2 1.2B on Dual DGX Spark report |
| DGX Station GB300 | 8-bit | 125K | 1,426 tok/s | 0.1 s | PDF: LFM2 1.2B on DGX Station GB300 report |
| 2× RTX PRO 6000 Blackwell | 8-bit | 125K | 737 tok/s | 0.1 s | PDF: LFM2 1.2B on 2× RTX PRO 6000 Blackwell report |
| 4× RTX PRO 6000 Blackwell | 8-bit | 125K | 737 tok/s | 0.1 s | PDF: LFM2 1.2B on 4× RTX PRO 6000 Blackwell report |
| 8× RTX PRO 6000 Blackwell | 8-bit | 125K | 737 tok/s | 0.1 s | PDF: LFM2 1.2B on 8× RTX PRO 6000 Blackwell report |
| A100 80 GB SXM | 8-bit | 125K | 799 tok/s | 0.1 s | PDF: LFM2 1.2B on A100 80 GB SXM report |
| H100 SXM | 8-bit | 125K | 1,053 tok/s | 0.1 s | PDF: LFM2 1.2B on H100 SXM report |
| 4× H200 NVL | 8-bit | 125K | 1,238 tok/s | 0.1 s | PDF: LFM2 1.2B on 4× H200 NVL report |
| 8× H200 NVL | 8-bit | 125K | 1,238 tok/s | 0.1 s | PDF: LFM2 1.2B on 8× H200 NVL report |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- llama.cpp
llama-server -hf LiquidAI/LFM2-1.2B-GGUF- vLLM
vllm serve LiquidAI/LFM2-1.2B- MLX (Mac)
mlx_lm.chat --model mlx-community/LFM2-1.2B-4bit
Questions
How much VRAM does LFM2 1.2B need?
LFM2 1.2B needs 677 MB for its weights at 4-bit, 1.5 GB with an 8K context and 2.3 GB with 32K. At 8-bit, 2.0 GB.
What is the minimum VRAM for LFM2 1.2B?
About 1.4 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the RTX 4060.
How big is LFM2 1.2B?
1.2B parameters: 2.2 GB of weights at 16-bit, 677 MB at 4-bit.
Can LFM2 1.2B run on an RTX 4090?
Yes. At 8-bit it holds 125K of context and writes about 490 tokens a second.