---

license: apache-2.0

base_model: KoarAI/LFM2.5-350M-Thinking-0003

tags:

• gguf
• llama.cpp
• ollama
• reasoning
• thinking
• cot
• liquid
• lfm
• full-finetune
• agentic
• mmlu-pro
• qwen
• koarai

language:

• en
• ru

pipeline_tag: text-generation

---

<div align="center">

<img src="https://huggingface.co/KoarAI/LFM2.5-350M-Thinking-0003-GGUF/resolve/main/banner.png" alt="KoarAI LFM2.5-350M Thinking 0003 GGUF Banner" width="100%" style="border-radius: 12px; box-shadow: 0 4px 20px rgba(0,0,0,0.3);"/>

🐨 KoarAI / LFM2.5-350M-Thinking-0003-GGUF






</div>

> [!NOTE]

> This is the official GGUF quantized release of KoarAI/LFM2.5-350M-Thinking-0003 for local edge inference with llama.cpp, Ollama, LM Studio, Jan, and other compatible inference engines.

---

📌 Release Note: Model Revision 0003

> [!IMPORTANT]

> Model Revision: 0003

> Key Architecture & Dataset Improvements:

> - Anti-Overfitting Training Policy: Trained with a calibrated 2.1 epochs limit with a cosine learning rate scheduler (lr=2.5e-5), preventing catastrophic forgetting and repetition loops while preserving pristine generation fluency.

> - Multi-Teacher Distillation Mix (~1,550 hand-crafted samples):

> 1. r0b0tlab/qwen3.8-max-distillation-50k (quality_score >= 6): Deep mathematical, algorithmic, and code reasoning distilled directly from Qwen 3.8 Max.

> 2. open-thoughts/OpenThoughts-114k: Step-by-step verification and multi-hop Chain-of-Thought traces encapsulated in <think> blocks.

> 3. TIGER-Lab/MMLU-Pro: Complex STEM, business, logic, and law benchmark multiple-choice reasoning.

> 4. Vikhrmodels/GrandMaster-PRO-MAX: High-quality natural Russian conversational mastery and instruction following.

---

🌟 Overview

KoarAI/LFM2.5-350M-Thinking-0003 is an ultra-compact, high-efficiency hybrid reasoning language model featuring native Chain-of-Thought (CoT) thinking capabilities.

Built upon the state-of-the-art Liquid Foundation Model architecture (LiquidAI/LFM2.5-350M), this model underwent 100% Full Parameter Fine-Tuning in bfloat16 precision. Despite having only 350 Million parameters, it produces structured internal step-by-step logic inside native <think> ... </think> blocks before delivering clean, concise final answers.

---

💾 Available GGUF Files

| Filename | Quantization | Size | Recommended Use |

|---|---|---|---|

| LFM2.5-350M-Thinking-0003-f16.gguf | FP16 | ~700 MB | Maximum precision and fidelity |

---

💡 Native Thinking Mode Example

<|im_start|>user
Сколько букв 'р' в слове 'программирование'?<|im_end|>
<|im_start|>assistant
<think>
1. Слово: 'программирование'.
2. Разберем по буквам: п-р-о-г-р-а-м-м-и-р-о-в-а-н-и-е.
3. Позиции буквы 'р':
   - 2-я буква: 'р' (п-Р-о...)
   - 5-я буква: 'р' (...г-Р-а...)
   - 10-я буква: 'р' (...и-Р-о...)
4. Итого: 3 буквы 'р'.
</think>
В слове «программирование» содержится **3** буквы «р».<|im_end|>

---

⚡ Quickstart with llama.cpp

# Run CLI
llama-cli -m LFM2.5-350M-Thinking-0003-f16.gguf -p "<|im_start|>user\nSolve: 32 + 32 - 42<|im_end|>\n<|im_start|>assistant\n<think>\n" -n 512 --temp 0.6

---

🐨 Maintained by KoarAI Lab

Released for the open-source AI community by KoarAI.