How much VRAM does LFM2.5 24B A2B need?
LFM2.5 24B A2B needs 13.5 GB for its weights at 4-bit, 14.4 GB with an 8K context and 14.9 GB with 32K. At 8-bit, 24.6 GB.
- Parameters
- 24B
- Active per token
- 2B
- Longest context
- 128K
- Released
- Oct 2026
- License
- LFM Open License v1.0
- Strongest at
- Agents
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 44.6 GB | 45.6 GB | 46.0 GB | 47.9 GB |
| 8-bit | 23.7 GB | 24.6 GB | 25.1 GB | 27.0 GB |
| 6-bit | 18.3 GB | 19.2 GB | 19.7 GB | 21.6 GB |
| 5-bit | 15.9 GB | 16.8 GB | 17.3 GB | 19.1 GB |
| 4-bit | 13.5 GB | 14.4 GB | 14.9 GB | 16.8 GB |
| 3-bit | 10.9 GB | 11.8 GB | 12.3 GB | 14.2 GB |
| 2-bit | 8.4 GB | 9.3 GB | 9.8 GB | 11.6 GB |
| 1.6-bit | 4.6 GB | 5.5 GB | 6.0 GB | 7.9 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | Too big | |||
| RTX 4060 Ti 16 GB | 4-bit | 44K | 119 tok/s | 0.6 s |
| RTX 3090 | 6-bit | 128K | 204 tok/s | 0.4 s |
| RTX 4090 | 6-bit | 128K | 212 tok/s | 0.2 s |
| RTX 5090 | 8-bit | 128K | 244 tok/s | 0.1 s |
| 2× RTX 3090 | 8-bit | 128K | 177 tok/s | 0.2 s |
| RTX PRO 6000 Blackwell | 8-bit | 128K | 244 tok/s | 0.1 s |
| M4 Max 128 GB | 8-bit | 128K | 86 tok/s | 2.2 s |
| M5 Max 128 GB | 8-bit | 128K | 92 tok/s | 0.6 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 128K | 57 tok/s | 1.3 s |
| DGX Spark 128 GB | 8-bit | 128K | 74 tok/s | 0.5 s |
| M3 Ultra 256 GB | 8-bit | 128K | 104 tok/s | 1.2 s |
| M5 Ultra 512 GB | 8-bit | 128K | 120 tok/s | 0.4 s |
| H100 SXM | 8-bit | 128K | 302 tok/s | 0.1 s |
| DGX Station GB300 | 8-bit | 128K | 354 tok/s | 0.1 s |
| 8× H200 NVL | 8-bit | 128K | 330 tok/s | 0.1 s |
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
- Ollama
ollama run lfm2.5:24b- llama.cpp
llama-server -hf unsloth/LFM2.5-24B-A2B-GGUF- vLLM
vllm serve LiquidAI/LFM2.5-24B-A2B- MLX (Mac)
mlx_lm.chat --model mlx-community/LFM2.5-24B-A2B-OptiQ-4bit
Questions
How much VRAM does LFM2.5 24B A2B need?
LFM2.5 24B A2B needs 13.5 GB for its weights at 4-bit, 14.4 GB with an 8K context and 14.9 GB with 32K. At 8-bit, 24.6 GB.
Can LFM2.5 24B A2B run on an RTX 4090?
Yes. At 6-bit it holds 128K of context and writes about 212 tokens a second.