@vivmarquez: Not every AI problem is a generation problem. For routing, classification, retrieval, policy checks, and similar tasks,…
Summary
Liquid AI releases LFM2.5-Encoder-230M and LFM2.5-Encoder-350M, bidirectional encoders optimized for non-generative tasks like classification and retrieval, offering fast CPU inference at long context.
View Cached Full Text
Cached at: 07/30/26, 01:48 AM
Not every AI problem is a generation problem.
For routing, classification, retrieval, policy checks, and similar tasks, generating an answer token by token can add unnecessary latency, cost, and complexity.
A fine-tuned encoder can handle many of these tasks directly: one forward pass, structured outputs, predictable latency, and no free-form response to parse.
That’s why I’m so excited about the new LFM2.5-Encoders.
Liquid AI (@liquidai): Today we release LFM2.5-Encoder-230M and LFM2.5-Encoder-350M: bidirectional encoders that stay fast at long context, even on CPU.
> LFM2.5-Encoder-230M: about 3.7x faster than ModernBERT-base on CPU at 8,192 tokens. Under 30s per forward pass, versus over a minute and a half. >
Similar Articles
LiquidAI/LFM2.5-Encoder-350M
Liquid AI releases LFM2.5-Encoder-350M, a multilingual bidirectional encoder built on the LFM2 architecture, offering strong quality for its size, 8k context, and efficient on-device performance across 15 languages.
LFM2.5-Encoders for Fast Long-Context Inference on CPU
Liquid AI releases LFM2.5-Encoders (230M and 350M), efficient encoder models optimized for long-context inference on CPU, matching or beating larger encoders on benchmarks with 3.7x speedup over ModernBERT-base.
@maximelabonne: We just released two new encoder models (MLM) in 2026 They're super fast, easy to train, and strongly multilingual. Try…
Liquid AI released two new encoder models, LFM2.5-Encoder-230M and LFM2.5-Encoder-350M, that are fast, easy to train, and strongly multilingual, with speed benchmarks showing over 3.7x improvement on CPU compared to ModernBERT-base.
LiquidAI/LFM2.5-230M
Liquid AI released LFM2.5-230M, a compact 230M-parameter hybrid model optimized for on-device deployment with fast edge inference speeds (213 tok/s on Galaxy S25 Ultra) and built for agentic tasks via reinforcement learning.
LiquidAI/LFM2.5-Embedding-350M
Liquid AI releases LFM2.5-Embedding-350M, a dense bi-encoder for multilingual retrieval supporting 11 languages, as a drop-in replacement for RAG pipelines.