SSyncDev

LFM3-Audio-2B

SyncDev model directory · Profile updated Nov 17, 2026

Compare with another model

LFM3-Audio-2B is Liquid AI's open 2B speech model: ASR, TTS, voice chat and interleaved text-and-speech output from one local checkpoint.

Key specifications

Specification score
35.0
Context window
32.77K
License
LFM Open License v1.0

The specification score summarizes recorded features and limits. It is not a benchmark result.

Pricing conditions: Open weights under LFM Open License v1.0; no hosted price confirmed.

Inputs and outputs

Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.

Accepts

  • Text input
  • Audio input

Produces

  • Text output
  • Audio output

Capabilities

Function callingAudioOpen weights

What is LFM3-Audio-2B?

Text, speech, or both at once. LFM3-Audio-2B can answer a spoken question with interleaved output, streaming words and 24 kHz audio from the same model, which removes the separate recogniser and synthesiser most voice apps chain together. Liquid AI released it on 27 October 2026 as the audio member of the LFM3 family, aimed at laptops and edge devices.

Its Hugging Face repo existed from September and went public with the launch.

Modes in one checkpoint

  • Speech recognition, with language set in the system prompt or inferred.
  • Live ASR in roughly 0.96-second chunks while someone talks. English only.
  • Speech synthesis, with a language and a female or male voice set in the system prompt.
  • Voice chat with interleaved text and speech. English only.
  • Speech-to-text chat, optionally reasoning first.
  • Audio understanding: questions about speech, music and sound events.
  • Translation and tool calling from spoken or written requests.

The English-only restriction on live ASR and voice chat is the detail to catch early if you serve other markets.

Under the hood

A causal audio encoder (sliding-window attention plus short convolutions, 16 kHz input) sits on the LFM3 backbone. Output speech goes through LiquidCodec, Liquid's tokenizer, at 12 tokens per 80 ms with streaming decode. LiquidCodec decodes at about 87x real time on a single CPU thread of an Apple M4 Max. Time to first token can be as low as 50 ms with streaming input. Context is 32K tokens, input plus output.

Reported results

Benchmark Score Note
English ASR WER 6.2 Voxtral Mini 3B 2507 is better at 5.5
Multilingual ASR WER/CER 6.0 Best in Liquid's table
English TTS UTMOSv2 3.8 Best in Liquid's table
VoiceBench 73.1 Best in Liquid's table
Tool calling (internal) 58.5 Gemma 4 E4B leads at 72.4

Liquid's own figures. Notice that English ASR is essentially flat against LFM2.5-Audio-1.5B (6.1); the gains are in conversation and synthesis.

Language count

Liquid's launch post says 18 languages. The model's speech list has 19 entries because it counts British and American English separately. Same thing, different counting. Research elsewhere put it at 19; the 18-language figure is the cleaner one.

Running it

BF16 for vLLM, SGLang and Transformers (5.20.2+, with trust_remote_code=True, plus librosa and soundfile), GGUF for llama.cpp, MLX 4-bit through MLX-VLM, and ONNX. Liquid does not recommend it for agentic coding or complex multi-step reasoning.

A voice-app trial

Record thirty realistic spoken requests in a noisy room. Measure end-of-speech to first audio, check that interleaved answers stay coherent when text and audio diverge, and have a person rate the voice. Repeat the ASR portion in every language you need, since voice chat will not help outside English.

See also audio-to-audio models and Liquid's provider page.

How to evaluate LFM3-Audio-2B

Compare the model on the work you need to complete.

Tasks and inputs

LFM3-Audio-2B is listed for Audio-to-audio, Automatic speech recognition, Text-to-speech. It accepts text, audio and produces text, audio. Use the publisher’s input format and preprocessing for your chosen deployment.

Capacity and cost

Documented limits: 32.77K context. Compare pronunciation, voice consistency and time to the first playable audio. Check the service’s billing unit for the actual script.

Choosing a deployment

The weights are downloadable under LFM Open License v1.0. Check runtime support and size the complete model and working context for your expected concurrency.

Frequently asked questions

Text, 24 kHz speech, or interleaved text and speech from the same model.

These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.

rime

Prism

Rime Prism is a speech-to-speech model for enterprise voice agents that settles each reply as inspectable text first, with parallel tool calls.

— context

Alibaba

Qwen-Audio-3.1-Realtime Flash

Qwen-Audio-3.1-Realtime Flash is Alibaba's lighter full-duplex speech model with audio and text in and out, function calling, web search and a 262K context.

262.14K context

ElevenLabs

Eleven Realtime v1 Max

Eleven Realtime v1 Max is ElevenLabs' reasoning speech-to-speech agent for support and multi-step tool workflows, OpenAI Realtime-compatible.

— context

ElevenLabs

Eleven Realtime v1 Mini

Eleven Realtime v1 Mini is ElevenLabs' fast, low-cost speech-to-speech voice agent stack, with an OpenAI Realtime-compatible API and 90+ languages.

— context

xAI

Grok Voice 3.0

Grok Voice 3.0 is xAI's speech-to-speech model for real-time voice agents, replacing Grok Voice 2.0 at the same $0.08-per-minute price.

— context

Amazon

Nova 2.5 Sonic

Nova 2.5 Sonic is Amazon's speech-to-speech model on Bedrock for real-time voice agents, with async tool calling, polyglot voices and a 256K context.

256K context