D1 2B VRAM requirements
| 4-bit weights: 1.1 GB | With 8K context: 1.9 GB | With 32K context: 2.2 GB | 8-bit, 8K context: 2.8 GB |
- Parameters
- 1.9B
- Longest context
- 125K
- Released
- Oct 2026
- License
- Lfm1.0
- Strongest at
- Text and images
Its licence (Lfm1.0) is not a standard open-source one: check the terms before commercial use.
How much VRAM D1 2B needs
| Precision | Weights | With 8K | With 32K | With 125K |
|---|---|---|---|---|
| 16-bit | 3.6 GB | 4.4 GB | 4.7 GB | 5.8 GB |
| 8-bit | 1.9 GB | 2.8 GB | 3.0 GB | 4.1 GB |
| 6-bit | 1.5 GB | 2.3 GB | 2.6 GB | 3.7 GB |
| 5-bit | 1.3 GB | 2.1 GB | 2.4 GB | 3.5 GB |
| 4-bit | 1.1 GB | 1.9 GB | 2.2 GB | 3.3 GB |
| 3-bit | 894 MB | 1.7 GB | 2.0 GB | 3.1 GB |
| 2-bit | 686 MB | 1.5 GB | 1.8 GB | 2.9 GB |
| 1.6-bit | 377 MB | 1.2 GB | 1.5 GB | 2.6 GB |
One conversation, 16-bit context cache, runtime overhead included.
GPUs and Macs that run D1 2B
| Machine | Best fit | Context | Speed | First reply | Report |
|---|---|---|---|---|---|
| RTX 4060 | 8-bit | 125K | 88 tok/s | 0.2 s | PDF: D1 2B on RTX 4060 report |
| RTX 5060 | 8-bit | 125K | 139 tok/s | 0.2 s | PDF: D1 2B on RTX 5060 report |
| RTX 3070 | 8-bit | 125K | 139 tok/s | 0.2 s | PDF: D1 2B on RTX 3070 report |
| RTX 3080 10 GB | 8-bit | 125K | 220 tok/s | 0.1 s | PDF: D1 2B on RTX 3080 10 GB report |
| RTX 3060 12 GB | 8-bit | 125K | 114 tok/s | 0.3 s | PDF: D1 2B on RTX 3060 12 GB report |
| Arc B580 | 8-bit | 125K | 96 tok/s | 0.2 s | PDF: D1 2B on Arc B580 report |
| RTX 4070 | 8-bit | 125K | 154 tok/s | 0.1 s | PDF: D1 2B on RTX 4070 report |
| RTX 4070 Super | 8-bit | 125K | 154 tok/s | 0.1 s | PDF: D1 2B on RTX 4070 Super report |
| RTX 5070 | 8-bit | 125K | 198 tok/s | 0.1 s | PDF: D1 2B on RTX 5070 report |
| RTX 4060 Ti 16 GB | 8-bit | 125K | 92 tok/s | 0.2 s | PDF: D1 2B on RTX 4060 Ti 16 GB report |
| RTX 5060 Ti 16 GB | 8-bit | 125K | 139 tok/s | 0.2 s | PDF: D1 2B on RTX 5060 Ti 16 GB report |
| Arc A770 16 GB | 8-bit | 125K | 115 tok/s | 0.2 s | PDF: D1 2B on Arc A770 16 GB report |
| Radeon RX 9070 XT | 8-bit | 125K | 154 tok/s | 0.1 s | PDF: D1 2B on Radeon RX 9070 XT report |
| RTX 4070 Ti Super | 8-bit | 125K | 198 tok/s | 0.1 s | PDF: D1 2B on RTX 4070 Ti Super report |
| RTX 4080 Super | 8-bit | 125K | 214 tok/s | 0.1 s | PDF: D1 2B on RTX 4080 Super report |
| RTX 5070 Ti | 8-bit | 125K | 252 tok/s | 0.1 s | PDF: D1 2B on RTX 5070 Ti report |
| RTX 5080 | 8-bit | 125K | 266 tok/s | 0.1 s | PDF: D1 2B on RTX 5080 report |
| Radeon RX 7900 XTX | 8-bit | 125K | 211 tok/s | 0.2 s | PDF: D1 2B on Radeon RX 7900 XTX report |
| RTX 3090 | 8-bit | 125K | 261 tok/s | 0.1 s | PDF: D1 2B on RTX 3090 report |
| RTX 4090 | 8-bit | 125K | 277 tok/s | 0.1 s | PDF: D1 2B on RTX 4090 report |
| Radeon AI PRO R9700 | 8-bit | 125K | 154 tok/s | 0.1 s | PDF: D1 2B on Radeon AI PRO R9700 report |
| RTX 5090 | 8-bit | 125K | 426 tok/s | 0.1 s | PDF: D1 2B on RTX 5090 report |
| 2× RTX 3090 | 8-bit | 125K | 261 tok/s | 0.1 s | PDF: D1 2B on 2× RTX 3090 report |
| 2× RTX 4090 | 8-bit | 125K | 277 tok/s | 0.1 s | PDF: D1 2B on 2× RTX 4090 report |
| RTX PRO 5000 Blackwell 48 GB | 8-bit | 125K | 346 tok/s | 0.1 s | PDF: D1 2B on RTX PRO 5000 Blackwell 48 GB report |
| RTX 6000 Ada | 8-bit | 125K | 266 tok/s | 0.1 s | PDF: D1 2B on RTX 6000 Ada report |
| L40S | 8-bit | 125K | 244 tok/s | 0.1 s | PDF: D1 2B on L40S report |
| 2× RTX 5090 | 8-bit | 125K | 426 tok/s | 0.1 s | PDF: D1 2B on 2× RTX 5090 report |
| 4× RTX 3090 | 8-bit | 125K | 261 tok/s | 0.1 s | PDF: D1 2B on 4× RTX 3090 report |
| RTX PRO 6000 Blackwell | 8-bit | 125K | 426 tok/s | 0.1 s | PDF: D1 2B on RTX PRO 6000 Blackwell report |
| Mac mini M4 16 GB | 8-bit | 125K | 44 tok/s | 2.4 s | PDF: D1 2B on Mac mini M4 16 GB report |
| Mac mini M4 24 GB | 8-bit | 125K | 44 tok/s | 2.4 s | PDF: D1 2B on Mac mini M4 24 GB report |
| Mac mini M4 32 GB | 8-bit | 125K | 44 tok/s | 2.4 s | PDF: D1 2B on Mac mini M4 32 GB report |
| Mac mini M4 Pro 24 GB | 8-bit | 125K | 91 tok/s | 1.2 s | PDF: D1 2B on Mac mini M4 Pro 24 GB report |
| Mac mini M4 Pro 48 GB | 8-bit | 125K | 91 tok/s | 1.2 s | PDF: D1 2B on Mac mini M4 Pro 48 GB report |
| Mac mini M4 Pro 64 GB | 8-bit | 125K | 91 tok/s | 1.2 s | PDF: D1 2B on Mac mini M4 Pro 64 GB report |
| M4 Max 128 GB | 8-bit | 125K | 145 tok/s | 0.6 s | PDF: D1 2B on M4 Max 128 GB report |
| M5 Max 128 GB | 8-bit | 125K | 158 tok/s | 0.2 s | PDF: D1 2B on M5 Max 128 GB report |
| M3 Ultra 96 GB | 8-bit | 125K | 174 tok/s | 0.3 s | PDF: D1 2B on M3 Ultra 96 GB report |
| M3 Ultra 256 GB | 8-bit | 125K | 174 tok/s | 0.3 s | PDF: D1 2B on M3 Ultra 256 GB report |
| M3 Ultra 512 GB | 8-bit | 125K | 174 tok/s | 0.3 s | PDF: D1 2B on M3 Ultra 512 GB report |
| M5 Ultra 512 GB | 8-bit | 125K | 221 tok/s | 0.1 s | PDF: D1 2B on M5 Ultra 512 GB report |
| Ryzen AI Max+ 395 64 GB | 8-bit | 125K | 69 tok/s | 0.3 s | PDF: D1 2B on Ryzen AI Max+ 395 64 GB report |
| Ryzen AI Max+ 395 128 GB | 8-bit | 125K | 69 tok/s | 0.3 s | PDF: D1 2B on Ryzen AI Max+ 395 128 GB report |
| RTX Spark | 8-bit | 125K | 96 tok/s | 0.1 s | PDF: D1 2B on RTX Spark report |
| DGX Spark | 8-bit | 125K | 88 tok/s | 0.1 s | PDF: D1 2B on DGX Spark report |
| Dual DGX Spark | 8-bit | 125K | 88 tok/s | 0.1 s | PDF: D1 2B on Dual DGX Spark report |
| DGX Station GB300 | 8-bit | 125K | 885 tok/s | 0.1 s | PDF: D1 2B on DGX Station GB300 report |
| 2× RTX PRO 6000 Blackwell | 8-bit | 125K | 426 tok/s | 0.1 s | PDF: D1 2B on 2× RTX PRO 6000 Blackwell report |
| 4× RTX PRO 6000 Blackwell | 8-bit | 125K | 426 tok/s | 0.1 s | PDF: D1 2B on 4× RTX PRO 6000 Blackwell report |
| 8× RTX PRO 6000 Blackwell | 8-bit | 125K | 426 tok/s | 0.1 s | PDF: D1 2B on 8× RTX PRO 6000 Blackwell report |
| A100 80 GB SXM | 8-bit | 125K | 466 tok/s | 0.1 s | PDF: D1 2B on A100 80 GB SXM report |
| H100 SXM | 8-bit | 125K | 629 tok/s | 0.1 s | PDF: D1 2B on H100 SXM report |
| 4× H200 NVL | 8-bit | 125K | 754 tok/s | 0.1 s | PDF: D1 2B on 4× H200 NVL report |
| 8× H200 NVL | 8-bit | 125K | 754 tok/s | 0.1 s | PDF: D1 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/d1-2B-GGUF- vLLM
vllm serve LiquidAI/d1-2B- MLX (Mac)
mlx_lm.chat --model mlx-community/d1-2B-OptiQ-4bit
Questions
How much VRAM does D1 2B need?
D1 2B needs 1.1 GB for its weights at 4-bit, 1.9 GB with an 8K context and 2.2 GB with 32K. At 8-bit, 2.8 GB.
What is the minimum VRAM for D1 2B?
About 1.7 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 D1 2B?
1.9B parameters: 3.6 GB of weights at 16-bit, 1.1 GB at 4-bit.
Can D1 2B run on an RTX 4090?
Yes. At 8-bit it holds 125K of context and writes about 277 tokens a second.