Laya: open-source non-autoregressive decision engine (Apache 2.0)
A 421M ModernBERT-based decision model that returns typed choices, scores, or yes/no judgments in a single forward pass (about 33-40ms on a T4), with a 322M multilingual variant and a Jev-compatible server.
Why it matters
The open challenger in the decision-model wave: pip install laya, 31K+ GitHub stars, 121 contributors, still trending on Hugging Face weeks after launch, a fine-tunable fast base instead of a closed API.
What you could build with it
An indie developer could self-host Laya as a sub-50ms decision gate in front of a larger model, screening prompt injections, routing support tickets, or triaging content, with no per-call API fees.
Does it hold up?
Substantive evidence: 31K+ stars, active contributors, and real deployments like the LinkedIn detector extension; but the headline 0.766 accuracy requires fine-tuning (zero-shot base is 0.362), so treat it as a fast base to specialize, not a zero-shot engine.
Built with Laya
- Open Source Is Pulling Away From JEV as LAYA Hits 121 Contributorsstartupfortune.com · Analysis of Laya's momentum: 30K+ stars, 121 contributors, 188 PRs merged in a week, three checkpoints, a Jev-compatible server, CLI, LangChain/LlamaIndex plugins and an MCP server.
- AI-Generated LinkedIn Post Detector Built with Laya Decision Enginelinkedin · A builder fine-tuned Laya and deployed it via Modal/FastAPI inside a Chrome extension that flags AI-generated LinkedIn posts in real time while scrolling.
- Laya Is INSANE: Open-Source AI Just Challenged Jevyoutube · Video walkthrough of Laya's architecture and community ports (MLX, CoreML, WASM, Android), with head-to-head benchmark numbers against the closed decision model.
Learn more
First spotted on hf: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.