Is LFM3-Audio-2B state of the art?Pending benchmarks.Announced Oct 27, 2026; no independent results yet.
Liquid AI's LFM3 audio model for laptops and edge devices. It takes text or audio and answers in text, speech or both: voice chat, speech recognition, synthesis and translation in 18 languages, audio understanding, and tool calls from spoken requests. Open weights under the LFM Open License v1.0.
The details so far
- Lab
- Liquid AI
- Announced
- Oct 27, 2026
- API
- Live · Liquid AI API and Playground, per Liquid's blog
- Open weights
- Published
- Parameters
- 2B
- Architecture
- LFM3 backbone, in-house causal audio encoder (16 kHz) and LiquidCodec audio tokenizer (24 kHz speech output)
- Input / output
- Text or audio in; text, speech, or interleaved text and speech out; streaming
- Context
- 32K tokens (input + output)
- Speech languages
- Arabic, Danish, German, Greek, English (British and American), Finnish, French, Hebrew, Italian, Japanese, Korean, Dutch, Polish, Portuguese, Spanish, Swedish, Turkish, Chinese
- License
- LFM Open License v1.0
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: better speech quality than LFM2.5-Audio, with English TTS at UTMOSv2 3.8 and 1.4% WER vs 3.2 and 1.9% for LFM2.5-Audio-1.5B, and time to first token as low as 50 ms.
Sources
LABLiquid AI: Introducing LFM3liquid.ai/blog/lfm3 · Oct 27, 2026REPOLiquidAI/LFM3-Audio-2B on Hugging Facehuggingface.co/LiquidAI/LFM3-Audio-2BTracking continues until an independent board lists it.