dlasher/LFM2.5-24B-A2B-GGUF

dlasher/ LFM2.5-24B-A2B-GGUF

15 Downloads Updated yesterday

pulled as BF16, requant to IQ4_NL, with combined_en_huge as dataset (English + Coding focus)

ollama run dlasher/LFM2.5-24B-A2B-GGUF:IQ4_NL

curl http://localhost:11434/api/chat \
  -d '{
    "model": "dlasher/LFM2.5-24B-A2B-GGUF:IQ4_NL",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='dlasher/LFM2.5-24B-A2B-GGUF:IQ4_NL',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'dlasher/LFM2.5-24B-A2B-GGUF:IQ4_NL',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size / Usage

Context

Input

LFM2.5-24B-A2B-GGUF:IQ4_NL

14GB · 128K context window · Text · yesterday

LFM2.5-24B-A2B-GGUF:IQ4_NL

14GB

128K

Text

Readme

PERPLEXITY QUALITY SCORECARD
 Base Model PPL      : 9.0464 (16-bit reference)
 Quantized Model PPL : 9.4549 (IQ4_NL target)
 Perplexity Drift    : +0.4085 (+4.52%)

LFM2.5-24B-A2B is a 24B-parameter reasoning model with only 2B active parameters, built for fast, private agents on-device. It is the largest and final model in the LFM2.5 family, which now spans two orders of magnitude: from LFM2.5-230M to LFM2.5-24B-A2B.

Best-in-class agent: Competitive with Qwen3.6-35B-A3B, a model with ~50% more total and active parameters, on agentic and tool-use tasks, scoring 48.09 on LiveClaw versus 38.19. Reasoning at MoE speed: Always thinks before it answers. With only 2B active parameters per token, each reasoning token is cheap.

Efficient inference: Runs up to 1.7× faster than comparable MoEs on-device, with 184 tok/s on an Apple M5 Max and up to 21K output tok/s on a single NVIDIA B300 GPU.

Find more information about LFM2.5-24B-A2B in our blog post. https://www.liquid.ai/blog/lfm2-5-24b-a2b

SOURCES : * https://www.liquid.ai/blog/lfm2-5-24b-a2b * https://huggingface.co/LiquidAI/LFM2.5-24B-A2B * https://huggingface.co/datasets/eaddario/imatrix-calibration

```
PERPLEXITY QUALITY SCORECARD
 Base Model PPL      : 9.0464 (16-bit reference)
 Quantized Model PPL : 9.4549 (IQ4_NL target)
 Perplexity Drift    : +0.4085 (+4.52%)
```

![image203-oOdUn61bDRSykk4H951SoDOHzqoOe1 (1).jpeg](/assets/dlasher/LFM2.5-24B-A2B-GGUF/edae62f4-c5e8-4ad5-b59c-c6516d1da340)

LFM2.5-24B-A2B is a 24B-parameter reasoning model with only 2B active parameters, built for fast, private agents on-device. It is the largest and final model in the LFM2.5 family, which now spans two orders of magnitude: from LFM2.5-230M to LFM2.5-24B-A2B.

Best-in-class agent: Competitive with Qwen3.6-35B-A3B, a model with ~50% more total and active parameters, on agentic and tool-use tasks, scoring 48.09 on LiveClaw versus 38.19.
Reasoning at MoE speed: Always thinks before it answers. With only 2B active parameters per token, each reasoning token is cheap.

Efficient inference: Runs up to 1.7× faster than comparable MoEs on-device, with 184 tok/s on an Apple M5 Max and up to 21K output tok/s on a single NVIDIA B300 GPU.

Find more information about LFM2.5-24B-A2B in our blog post. [https://www.liquid.ai/blog/lfm2-5-24b-a2b](url)

SOURCES : 
* https://www.liquid.ai/blog/lfm2-5-24b-a2b
* https://huggingface.co/LiquidAI/LFM2.5-24B-A2B
* https://huggingface.co/datasets/eaddario/imatrix-calibration

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)