Liquid AI's 74M-param LiquidCodec hits 361x realtime encoding on a single CPU thread
helloiamleonie · x · 2026-11-13
Liquid AI built its own 74M-parameter streaming audio codec, LiquidCodec, after third-party tokenizers became a bottleneck for LFM3-Audio-2B and LFM3-Omni-2B. Designed around predictable tokens and fast reconstruction, it runs at 361x realtime encoding and 87x decoding on a single CPU thread of an Apple M4 Max (FP32), making on-device audio AI practical. An engineering blog details the design.
Related event: Liquid AI Unveils 74M-Parameter Streaming Audio Codec LiquidCodec(3 posts)→
More from Infra
- SGLang Summit: agents now write ~95% of merged PRs as v1.0 and Miles v0.2 ship — hsu_byron · 2026-11-13
- Bittensor mapped: 128 subnets form a full-stack decentralized AI supply chain — markjeffrey · 2026-11-13
- One week of irony: data center blockades, an NVIDIA selloff, and economic jitters — Dan_Jeffries1 · 2026-11-13
- Nvidia author Tae Kim: only two choices to succeed Jensen Huang as CEO — firstadopter · 2026-11-13
- AI SOC monitoring agents emerge as best practice; KV cache optimization called critical to cyber safety — AccBalanced · 2026-11-13
- Marin 535B-A23B Architecture Postmortem: MoE, XSA, MuonH and 18T-Token Curves — StasBekman · 2026-11-13