Liquid AI has released two new open-source decision-making models, d1-3B and d1-omni-600M, which can be downloaded and run locally. The d1-3B model, based on LFM2.5-VL-3B, handles text and image inputs, while the d1-omni-600M, based on LFM2.5-Encoder-350M, supports text, image, and audio inputs. These models are designed for tasks like selecting appropriate options with confidence scores and are noted for their speed and cost-effectiveness compared to traditional LLMs and OpenAI's Decisions API. Early user reports indicate impressive performance on consumer hardware, with d1-3B achieving decision processing times as low as 25 milliseconds on a GeForce RTX 3060. AI
IMPACT These open-source decision models offer faster and cheaper alternatives to LLMs for specific tasks, potentially accelerating adoption in edge computing and specialized applications.
RANK_REASON Open-source release of new decision-making models by an AI lab. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Mastodon — mastodon.social →
- d1-3B
- d1-omni-600M
- Decisions API
- Dogukan
- GeForce RTX 3060
- Hugging Face
- LFM2.5-Encoder-350M
- LFM2.5-VL-3B
- Liquid AI
- OpenAI
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
