LFM2.5-2.6B
Liquid AI·
LLMsopen-weightMobile / NPU
Overview
On-device agentic model from Liquid AI, 2.69B parameters with a 131K-token context window and tool calling, trained to work inside real agent harnesses rather than to top knowledge benchmarks. Liquid measures 220 tok/s on an Apple M5 Max, 113 tok/s on an AMD Ryzen AI Max+ 395 and roughly 30 tok/s on a phone, all under 2.5GB of memory, and reports it beating the 4x larger Qwen3.5-9B on ToolSandbox (77.83 vs 76.44) and on every instruction-following benchmark. No MMLU or GPQA figures have been published, so the benchmark fields stay null.
Capabilities and innovations
Runs under 2.5GB memory, ~30 tok/s on a phone131K-token context windowTool calling for on-device agentsOpen weightsAgent-harness training at 2.6B scaleBeats a 4x larger model on tool-use benchmarks
Benchmarks
No benchmark values are recorded for this model.
- MMLU: not reported by the vendor
- MMLU-Pro: not reported by the vendor
- GPQA Diamond: not reported by the vendor
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Pro: not reported by the vendor
- Terminal-Bench 2.x: not reported by the vendor
Architecture and hardware
- Parameters
- 2.69B total (Hybrid)
- Estimated VRAM at Q4
- ~2 GB, Mobile / NPU class
- Quantization formats
- GGUF, INT4
- Recommended runtime
- llama.cpp
- License
- LFM Open License v1.0
Links
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.