This is a deterministic native-MLX conversion of LiquidAI's QAD-trained Q4_0 GGUF LFM2.5-1.2B-Instruct-QAD-Q4_0.gguf at immutable revision afbd8eaeab5dd94ba0b079ebfb02517d19641e38.

Every Q4_0 projection nibble and FP16 block scale is retained exactly and repacked into MLX affine 4-bit/group-32 tensors with bias = -8 * scale.
The single Q6_K tied token embedding is decoded and requantized to MLX affine 6-bit/group-64; its measured maximum and mean elementwise errors are recorded in MERERUN_CONVERSION.json.
MERERUN_CONVERSION.json records all pinned inputs and output hashes. The converter is scripts/model-conversion/convert_lfm25_qad_mlx.py in the public mere.run repository.
Liquid reports that QAD retains Q4_0's compact footprint and native Q4_0 throughput while recovering accuracy lost by post-training quantization. Their throughput points use llama.cpp because the QAD and PTQ checkpoints share that GGUF runtime path. This native MLX conversion preserves the QAD-trained projection values, but MLX throughput must be measured separately.
Liquid's original LFM Open License terms remain applicable.

shell

mere.run model pull text-chat-lfm25-1.2b-qad-4bit --accept-model-license
mere.run text chat --model text-chat-lfm25-1.2b-qad-4bit --stats --prompt "Hello"

Original model card

library_name: transformers license: other license_name: lfm1.0 license_link: LICENSE language:

en
ar
zh
fr
de
ja
ko
es pipeline_tag: text-generation tags:
liquid
lfm2.5
edge base_model: LiquidAI/LFM2.5-1.2B-Base

Try LFM • Docs • LEAP • Discord

LFM2.5-1.2B-Instruct

LFM2.5 is a new family of hybrid models designed for on-device deployment. It builds on the LFM2 architecture with extended pre-training and reinforcement learning.

Best-in-class performance: A 1.2B model rivaling much larger models, bringing high-quality AI to your pocket.
Fast edge inference: 239 tok/s decode on AMD CPU, 82 tok/s on mobile NPU. Runs under 1GB of memory with day-one support for llama.cpp, MLX, and vLLM.
Scaled training: Extended pre-training from 10T to 28T tokens and large-scale multi-stage reinforcement learning.

Find more information about LFM2.5 in our blog post.

🗒️ Model Details

Model	Parameters	Description
LFM2.5-1.2B-Base	1.2B	Pre-trained base model for fine-tuning
LFM2.5-1.2B-Instruct	1.2B	General-purpose instruction-tuned model
LFM2.5-1.2B-Thinking	1.2B	General-purpose reasoning model
LFM2.5-1.2B-JP	1.2B	Japanese-optimized chat model
LFM2.5-VL-1.6B	1.6B	Vision-language model with fast inference
LFM2.5-Audio-1.5B	1.5B	Audio-language model for speech and text I/O

LFM2.5-1.2B-Instruct is a general-purpose text-only model with the following features:

Number of parameters: 1.17B
Number of layers: 16 (10 double-gated convolution blocks + 6 GQA blocks)
Training budget: 28T tokens
Context length: 32,768 tokens
Vocabulary size: 65,536
Knowledge cutoff: Mid-2024
Languages: English, Arabic, Chinese, French, German, Japanese, Korean, Spanish
Generation parameters:
- temperature: 0.1
- top_k: 50
- repetition_penalty: 1.05

Model	Description
LFM2.5-1.2B-Instruct	Original model checkpoint in native format. Best for fine-tuning or inference with Transformers and vLLM.
LFM2.5-1.2B-Instruct-GGUF	Quantized format for llama.cpp and compatible tools. Optimized for CPU inference and local deployment with reduced memory usage.
LFM2.5-1.2B-Instruct-ONNX	ONNX Runtime format for cross-platform deployment. Enables hardware-accelerated inference across diverse environments (cloud, edge, mobile).
LFM2.5-1.2B-Instruct-MLX	MLX format for Apple Silicon. Optimized for fast inference on Mac devices using the MLX framework.

We recommend using it for agentic tasks, data extraction, and RAG. It is not recommended for knowledge-intensive tasks and programming.

Chat Template

LFM2.5 uses a ChatML-like format. See the Chat Template documentation for details. Example:

<|startoftext|><|im_start|>system
You are a helpful assistant trained by Liquid AI.<|im_end|>
<|im_start|>user
What is C. elegans?<|im_end|>
<|im_start|>assistant

You can use tokenizer.apply_chat_template() to format your messages automatically.

Tool Use

LFM2.5 supports function calling as follows:

Function definition: We recommend providing the list of tools as a JSON object in the system prompt. You can also use the tokenizer.apply_chat_template() function with tools.
Function call: By default, LFM2.5 writes Pythonic function calls (a Python list between <|tool_call_start|> and <|tool_call_end|> special tokens), as the assistant answer. You can override this behavior by asking the model to output JSON function calls in the system prompt.
Function execution: The function call is executed, and the result is returned as a "tool" role.
Final answer: LFM2 interprets the outcome of the function call to address the original user prompt in plain text.

See the Tool Use documentation for the full guide. Example:

<|startoftext|><|im_start|>system
List of tools: [{"name": "get_candidate_status", "description": "Retrieves the current status of a candidate in the recruitment process", "parameters": {"type": "object", "properties": {"candidate_id": {"type": "string", "description": "Unique identifier for the candidate"}}, "required": ["candidate_id"]}}]<|im_end|>
<|im_start|>user
What is the current status of candidate ID 12345?<|im_end|>
<|im_start|>assistant
<|tool_call_start|>[get_candidate_status(candidate_id="12345")]<|tool_call_end|>Checking the current status of candidate ID 12345.<|im_end|>
<|im_start|>tool
[{"candidate_id": "12345", "status": "Interview Scheduled", "position": "Clinical Research Associate", "date": "2023-11-20"}]<|im_end|>
<|im_start|>assistant
The candidate with ID 12345 is currently in the "Interview Scheduled" stage for the position of Clinical Research Associate, with an interview date set for 2023-11-20.<|im_end|>

🏃 Inference

LFM2.5 is supported by many inference frameworks. See the Inference documentation for the full list.

Name	Description	Docs	Notebook
Transformers	Simple inference with direct access to model internals.	Link	Colab link
vLLM	High-throughput production deployments with GPU.	Link	Colab link
llama.cpp	Cross-platform inference with CPU offloading.	Link	Colab link
MLX	Apple's machine learning framework optimized for Apple Silicon.	Link	—
LM Studio	Desktop application for running LLMs locally.	Link	—

Here's a quick start example with Transformers:

python

from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

model_id = "LiquidAI/LFM2.5-1.2B-Instruct"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    dtype="bfloat16",
#   attn_implementation="flash_attention_2" <- uncomment on compatible GPU
)
tokenizer = AutoTokenizer.from_pretrained(model_id)
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

prompt = "What is C. elegans?"

input_ids = tokenizer.apply_chat_template(
    [{"role": "user", "content": prompt}],
    add_generation_prompt=True,
    return_tensors="pt",
    tokenize=True,
).to(model.device)

output = model.generate(
    input_ids,
    do_sample=True,
    temperature=0.1,
    top_k=50,
    repetition_penalty=1.05,
    max_new_tokens=512,
    streamer=streamer,
)

🔧 Fine-Tuning

We recommend fine-tuning LFM2.5 for your specific use case to achieve the best results.

Name	Description	Docs	Notebook
CPT (Unsloth)	Continued Pre-Training using Unsloth for text completion.	Link	Colab link
CPT (Unsloth)	Continued Pre-Training using Unsloth for translation.	Link	Colab link
SFT (Unsloth)	Supervised Fine-Tuning with LoRA using Unsloth.	Link	Colab link
SFT (TRL)	Supervised Fine-Tuning with LoRA using TRL.	Link	Colab link
DPO (TRL)	Direct Preference Optimization with LoRA using TRL.	Link	Colab link
GRPO (Unsloth)	GRPO with LoRA using Unsloth.	Link	Colab link
GRPO (TRL)	GRPO with LoRA using TRL.	Link	Colab link

📊 Performance

Benchmarks

We compared LFM2.5-1.2B-Instruct with relevant sub-2B models on a diverse suite of benchmarks.

Model	GPQA	MMLU-Pro	IFEval	IFBench	Multi-IF	AIME25	BFCLv3
LFM2.5-1.2B-Instruct	38.89	44.35	86.23	47.33	60.98	14.00	49.12
Qwen3-1.7B (instruct)	34.85	42.91	73.68	21.33	56.48	9.33	46.30
Granite 4.0-1B	24.24	33.53	79.61	21.00	43.65	3.33	52.43
Llama 3.2 1B Instruct	16.57	20.80	52.37	15.93	30.16	0.33	21.44
Gemma 3 1B IT	24.24	14.04	63.25	20.47	44.31	1.00	16.64

GPQA, MMLU-Pro, IFBench, and AIME25 follow ArtificialAnalysis's methodology. For IFEval and Multi-IF, we report the average score across strict and loose prompt and instruction accuracies. For BFCLv3, we report the final weighted average score with a custom Liquid handler to support our tool use template.

Inference speed

LFM2.5-1.2B-Instruct offers extremely fast inference speed on CPUs with a low memory profile compared to similar-sized models.

In addition, we are partnering with AMD, Qualcomm, and Nexa AI to bring the LFM2.5 family to NPUs. These optimized models are available through our partners, enabling highly efficient on-device inference. The following numbers have been calculated using 1K prefill and 100 decode tokens:

Device	Inference	Framework	Model	Prefill (tok/s)	Decode (tok/s)	Memory (GB)
Qualcomm Snapdragon® X Elite	NPU	NexaML	LFM2.5-1.2B-Instruct	2591	63	0.9GB
Qualcomm Snapdragon® Gen4 (ROG Phone9 Pro)	NPU	NexaML	LFM2.5-1.2B-Instruct	4391	82	0.9GB
Qualcomm Snapdragon® Gen4 (Samsung Galaxy S25 Ultra)	CPU	llama.cpp (Q4_0)	LFM2.5-1.2B-Instruct	335	70	719MB
Qualcomm Snapdragon® Gen4 (Samsung Galaxy S25 Ultra)	CPU	llama.cpp (Q4_0)	Qwen3-1.7B	181	40	1306MB

These capabilities unlock new deployment scenarios across various devices, including vehicles, mobile devices, laptops, IoT devices, and embedded systems.

📬 Contact

Got questions or want to connect? Join our Discord community
If you are interested in custom solutions with edge deployment, please contact our sales team.

Citation

bibtex

@article{liquidai2025lfm2,
  title={LFM2 Technical Report},
  author={Liquid AI},
  journal={arXiv preprint arXiv:2511.23404},
  year={2025}
}

Sawfwair/LFM2.5-1.2B-Instruct-QAD-MLX-4bit

text-chat-lfm25-1.2b-qad-4bit