Liquid AI releases open-weight reasoning model with expanded context

Liquid AI released LFM2.5-24B-A2B on October 13, making base and post-trained checkpoints available on Hugging Face for agents designed to run on-device, Unite.AI reported. The reasoning model expands its predecessor's context window from 32,768 to 131,072 tokens, allowing it to process longer inputs. The release supports llama.cpp, MLX, vLLM, SGLang and ONNX. Hosted inference on Liquid Console costs $0.10 per million input tokens, $0.05 for cached input and $1.00 per million output tokens. Liquid AI reports 184 tokens per second on an Apple M5 Max with 15 GB of memory, measured using quantization and a 4K-token input context. Its comparison tables show larger models retaining an edge on knowledge-heavy and coding benchmarks.