Oriel — AI & technology worth noticing, every day.

LiquidAI/LFM3-Omni-2B

AI model · https://huggingface.co/LiquidAI/LFM3-Omni-2B

First seen by Oriel on .

AI summary

LFM3-Omni-2B is a 2.17B-parameter model that handles text, images, video, and audio in one system, producing text, speech, or both for on-device use.

  • It supports speech recognition, translation, text-to-speech, image and audio understanding, and video questions that combine visual frames with the soundtrack.
  • Inputs include 16 kHz audio and native-resolution images; video defaults to 2 fps and 4 to 384 frames, with frame limits trading context and memory against coverage.
  • Outputs include text, 24 kHz speech, or interleaved text and speech, with a 32K-token context and support for multiple text and speech languages.
  • BF16 checkpoints target GPU inference through SGLang or Transformers, while GGUF supports local CPU and GPU inference through llama.cpp.
  • The model is explicitly non-reasoning; the page recommends greedy decoding for text and specified temperature and repetition-penalty settings for speech.

#multimodal models #speech recognition #text-to-speech #video understanding

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Models: 711, latest observation ; +604 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Models: 22,083, latest observation ; +21,369 over observed period since

Aliases

SourceExternal IDURL
Hugging Face ModelsLiquidAI/LFM3-Omni-2Bhttps://huggingface.co/LiquidAI/LFM3-Omni-2B Read discussion