{"v":1,"id":"model:hf:LiquidAI/LFM2.5-2.6B-GGUF","slug":"model-liquidai-lfm2-5-2-6b-gguf","kind":"model","category":"llm","title":"LFM2.5-2.6B-GGUF","summary":"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.","source":{"provider":"hf","ref":"LiquidAI/LFM2.5-2.6B-GGUF","url":"https://huggingface.co/LiquidAI/LFM2.5-2.6B-GGUF","rev":"e7caca5d835a3901a8e0d63e94009429bafafdfc","fetchedAt":"2026-10-02T20:59:26.626Z","etag":"W/\"29c3-yU7aErCK/t5EaBf7bR1fU7x7FlA\""},"author":{"name":"LiquidAI","url":"https://huggingface.co/LiquidAI"},"license":{"spdx":null,"raw":"other","open":null,"note":"custom license: read it at the source before installing"},"metrics":{"downloads":1066051,"downloadsWeek":327838,"likes":360,"takenAt":"2026-10-02T20:59:26.626Z"},"tags":["gguf","safetensors","liquid","lfm2.5","llama.cpp","text-generation","ar","zh","en","fr","de","hi","id","it","ja","ko","pl","pt","ru","es","th","vi","endpoints_compatible","conversational"],"pipeline":"text-generation","links":{"github":"ggml-org/llama.cpp","npm":"node-llama-cpp"},"updatedAt":"2026-09-22T20:42:43.000Z","collectedAt":"2026-10-02T20:59:26.626Z","review":{"numbers":["1,066,051 downloads on Hugging Face","360 likes","license other","1.6 GB for LFM2.5-2.6B-Q4_K_M.gguf","327,838 npm downloads a week for node-llama-cpp","latest node-llama-cpp@3.22.1"],"log":null},"trust":"unlabeled","health":{"status":"alive","checkedAt":"2026-10-02T20:59:26.626Z","http":200},"description":"Try LFM •\n    Docs •\n    LEAP •\n    Discord\n\n# LFM2.5-2.6B-GGUF\n\nLFM2.5 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.\n\nFind more details in the original model card: https://huggingface.co/LiquidAI/LFM2.5-2.6B\n\n## 🏃 How to run LFM2\n\nExample usage with [llama.cpp](https://github.com/ggml-org/llama.cpp):\n\n```\nllama-cli -hf LiquidAI/LFM2.5-2.6B-GGUF --conversation \\\n    --temp 0.1 --top-k 50 --repeat-penalty 1.1\n```\n\n## QAD Q4_0 GGUF\n\nThe Quantization-Aware Distillation (QAD) checkpoint is available as\n[`LFM2.5-2.6B-QAD-Q4_0.gguf`](https://huggingface.co/LiquidAI/LFM2.5-2.6B-GGUF/blob/main/LFM2.5-2.6B-QAD-Q4_0.gguf).\n\nThis is distinct from the post-training-quantized `LFM2.5-2.6B-Q4_0.gguf`;\nboth use the GGUF Q4_0 format.\n\n## QAD source weights (safetensors)\n\nThe original FP32 QAD source checkpoint is available in\n[`qad/`](https://huggingface.co/LiquidAI/LFM2.5-2.6B-GGUF/tree/main/qad), with its model config,\ntokenizer, generation defaults, and the same chat template as the released QAD GGUF.\nIt can be loaded in Transformers by passing `subfolder=\"qad\"`:\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nrepo_id = \"LiquidAI/LFM2.5-2.6B-GGUF\"\ntokenizer = AutoTokenizer.from_pretrained(repo_id, subfolder=\"qad\")\nmodel = AutoModelForCausalLM.from_pretrained(\n    repo_id, subfolder=\"qad\", dtype=\"auto\", device_map=\"auto\"\n)\n\ninputs = tokenizer.apply_chat_template(\n    [{\"role\": \"user\", \"content\": \"What is 2 + 2?\"}],\n    tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors=\"pt\",\n).to(model.device)\noutputs = model.generate(**inputs, max_new_tokens=128)\nprint(tokenizer.decode(outputs[0, inputs[\"input_ids\"].shape[-1]:], skip_special_tokens=True))\n```\n\nThese weights are intended for fine-tuning and experimentation. Published QAD\nresults apply to the Q4_0 GGUF; direc…\n\nSource: https://huggingface.co/LiquidAI/LFM2.5-2.6B-GGUF","install":{"kind":"model","hfId":"LiquidAI/LFM2.5-2.6B-GGUF","gated":false,"format":"gguf","files":[{"name":"LFM2.5-2.6B-BF16.gguf","size":5403158528,"quant":"BF16","sha256":"590b1534d5ec57dbddb750d25468d7ba4df0d1976531a731590418efa360edb5"},{"name":"LFM2.5-2.6B-F16.gguf","size":5403158528,"quant":"F16","sha256":"e041c231351185eb390f9c417d3bfd1815869a50a8589f3f86e5b9add3c529f1"},{"name":"LFM2.5-2.6B-Q4_0.gguf","size":1593894912,"quant":"Q4_0","sha256":"e1a61bf937bc60726e18626e97f7ee9bfd2574d95744c2ed909de98b78006fbe"},{"name":"LFM2.5-2.6B-Q4_K_M.gguf","size":1674455040,"quant":"Q4_K_M","sha256":"02a8b7e17487d326e46d68ce0ba24211e1b80a14c4cd0597fa73c1cd697f52ed"},{"name":"LFM2.5-2.6B-Q5_K_M.gguf","size":1939744768,"quant":"Q5_K_M","sha256":"17ce54cc676e15e572adebfeeed8da6abd12b0f68d595be19b59e662476c21b3"},{"name":"LFM2.5-2.6B-Q6_K.gguf","size":2221615104,"quant":"Q6_K","sha256":"2e74b1a0979a4a1936a408445147d103b8f15b2e2ec31c65fa0166f9069c250d"},{"name":"LFM2.5-2.6B-Q8_0.gguf","size":2874779648,"quant":"Q8_0","sha256":"1e22128dfa128bdfb684da167e74e072d0a056baa7d06d9f280291e2839b0fc9"},{"name":"LFM2.5-2.6B-QAD-Q4_0.gguf","size":1593894944,"quant":"Q4_0","sha256":"a247afd6414918eac8e520a9e6137dc271235461ecbe1180462221d5b8d40b03"},{"name":"qad/model.safetensors","size":10788824456,"sha256":"b81e17164b0c05fc1dfcfc3bd2b39b8cb95c77c00b20f068880ce0e547a70be8"}],"totalBytes":33493525928,"suggestedFile":"LFM2.5-2.6B-Q4_K_M.gguf","requirements":{"ramGb":3,"diskBytes":1674455040,"note":"estimate: suggested file size × 1.15 + 0.5 GB; a real measurement comes with lsh models"},"runWith":["llama.cpp","ollama"],"command":"lsh models install hf:LiquidAI/LFM2.5-2.6B-GGUF"}}