LFM2.5-2.6B
A small-parameter language model developed by Liquid AI with approximately 2.6 billion parameters, suited for reinforcement learning training experiments in agent-oriented scenarios.
Timeline (last 90 days)
d1-3B has 3.12B parameters, its vision encoder is a shape-optimized version of SigLIP2 NaFlex (400M), its context length is 32,768 tokens, its vocabulary is 128,000, its weight format is Safetensors/BF16, its base model is LFM2.5-VL-3B (the text backbone was weight-averaged with LFM2.5-2.6B), and its license is lfm1.0.
Training simultaneously across 4 harnesses improved Liquid AI's LFM2.5-2.6B model score from 42% to 54%.
By providing a small reward for completing tasks in fewer steps, the number of tool calls was reduced by 31%.
All Facts (3)
d1-3B has 3.12B parameters, its vision encoder is a shape-optimized version of SigLIP2 NaFlex (400M), its context length is 32,768 tokens, its vocabulary is 128,000, its weight format is Safetensors/BF16, its base model is LFM2.5-VL-3B (the text backbone was weight-averaged with LFM2.5-2.6B), and its license is lfm1.0.
50%UnverifiedTraining simultaneously across 4 harnesses improved Liquid AI's LFM2.5-2.6B model score from 42% to 54%.
50%UnverifiedBy providing a small reward for completing tasks in fewer steps, the number of tool calls was reduced by 31%.
50%