Skip to main content
Across the library:
  • Context length depends on the model: 32K tokens for most text, vision, and audio models, 128K for LFM2.5-8B-A1B, 131K for LFM2.5-2.6B and LFM2.5-24B-A2B, 8K for the LFM2.5 encoders, and 16K for d1-omni-600M. Each model page lists its exact limit.
  • Text and vision models run with Transformers, llama.cpp, vLLM, SGLang, MLX, Ollama, and Atomic Chat. Audio, retrieval, encoder, and decision models support fewer runtimes (d1-3B and d1-omni-600M: Transformers, llama.cpp, and mlx-vlm); check each model page.
  • Most open-weight models can be fine-tuned with Liquid Finetune, TRL, or Unsloth. The Trainable? column in the matrix below lists the documented path for each model.
Start with the model family that matches your input and output shape, then choose a runtime based on where you want to run it. Use the complete matrix below when you need exact repository and format availability.

​
Model Families

Text Models

Chat, tool calling, structured output, and classification.

Vision Models

Image understanding with LFM backbones and custom encoders.

Audio Models

Interleaved audio/text models for TTS, ASR, and voice chat.

Decision Models

Purpose-built models for classification, routing, and scoring.

Liquid Nanos

Task-specific models for extraction, summarization, RAG, and translation.

​
Common Workflows

GPU Serving

Use vLLM or SGLang for high-throughput serving, and Transformers for direct Python inference.

Local and On-Device

Use llama.cpp, Ollama, Atomic Chat, or MLX depending on platform and packaging needs. To embed a model in an iOS, Android, or desktop app, see Build with llama.cpp.

Fine-Tuning

Start with Liquid Finetune for managed workflows, or use TRL and Unsloth for framework-level control.

Model Repositories

Browse LiquidAI collections on Hugging Face for model weights, GGUF exports, MLX packages, ONNX exports, and model cards.

​
Formats

Use the format that matches your runtime and deployment target. The matrix below links the exact repository for each model and format.
  • GGUF — Best for local CPU/GPU inference on any platform. Use with llama.cpp, LM Studio, Ollama, or Atomic Chat. Every open-weight model in the matrix has a LiquidAI/<model>-GGUF repository.
  • MLX — Best for Mac users with Apple Silicon. Leverages unified memory for fast inference via MLX or Atomic Chat. Many models have official LiquidAI/<model>-MLX-<bits> repositories; the others are community conversions on mlx-community.
  • ONNX — Best for production deployments and edge devices. Cross-platform with ONNX Runtime across CPUs, GPUs, and accelerators. Available as LiquidAI/<model>-ONNX for the main LFM2.5 text, vision, and audio models, and for some LFM2 models through onnx-community; see the matrix.

​
Quantization

Quantization reduces model size and speeds up inference with minimal quality loss. Available options by format:
  • GGUF — Text and vision repositories usually include Q4_0, Q4_K_M, Q5_K_M, Q6_K, Q8_0, BF16, and F16. Some repositories ship a smaller set (for example, d1-omni-600M has only Q8_0, BF16, and F16), so check the file list. Q4_K_M offers the best balance of size and quality.
  • MLX — Official LiquidAI repositories usually come in 4bit, 5bit, 6bit, 8bit, and bf16. Some ship a subset (the d1 models have 8bit and fp16, plus 4bit for d1-3B). 8bit is recommended.
  • ONNX — Supports FP32, FP16, Q4, and Q8 (some repositories also include Q4F16). Q4 is recommended for most deployments.

​
Complete Model Matrix

Looking for an older model? Deprecated models and their recommended replacements are listed on the Deprecations page.