Radar / AI infrastructure / laya
laya: calibrated classification layer for routing and guardrails
Compact text-classification model for routing, scoring, guardrails and moderation, trained for calibrated decisions.
Why it matters
Apache-2.0 'system-one' decision layer (RL-calibrated) that sits in front of LLMs as a fast guardrail/router; #1 trending on HuggingFace this week.
What you could build with it
Use as a cheap pre-filter to route radar ingest items into the right category before any LLM pass.
Does it hold up?
Strongest third-party signal of the four: builders are already wiring Laya into agent stacks (MCP server) and shipping it on-device (Core ML port with measured parity). Independent evals agree on the catch: it's a fast base to fine-tune, not a zero-shot decision engine.
Built with laya
- FluidInference/laya-coreml — independent Core ML port of laya-multilingualhuggingface · Independent Core ML conversion of laya-multilingual with 16-fixture parity checks against the PyTorch reference (16/16 argmax agreement, max prob error 0.0021) and 3.6ms/question on an Apple M5 Pro; drives a Swift Tetris demo played entirely by Laya decisions.
- AI Weekly: Convai ships Laya, a 421M ModernBERT decision modelaiweekly · Third-party analysis crediting the 32.8ms p50 latency vs TypeSafe Jev's 236-276ms, while flagging that the 0.766 accuracy headline needs fine-tuning on the benchmark's own split (zero-shot is 0.362, near random).
- nventimiglia/laya-mcpgithub · 5 ★ · Third-party MCP server exposing Laya's choice/score/noul primitives to agents, with measured evals vs TypeSafe Jev and Ollama Gemma4 on 80 fixtures (Laya ~5x faster than Jev, least accurate; strong on moderation/guardrails, weak on knowledge MCQs) plus a fine-tuning guide.
First spotted on huggingface: source.
More AI infrastructure
Modular open-sources MAX inference server, Mojo stdlib and accelerator kernelsThe core of Modular's unified AI deployment stack is now open: the MAX inference server (OpenAI-compatible endpoints)…infra · JEV 0.76OmniRouteFree MIT AI gateway: one endpoint, 290+ providers and 500+ models with quota-aware auto-fallback.infra · JEV 0.75Cloudflare Agents Week: Sandboxes GA, 50K concurrent Workflows, Managed OAuth for agentsA dozen agent-infrastructure launches in one week: persistent Linux Sandboxes (GA) with real shell/filesystem/state…infra · JEV 0.69ai-memory: long-term memory for agent coding CLIsRust solution for long-term memory for agent coding CLIs, facilitating handoff between different agents and sessions.infra · JEV 0.68DeepSeek open-sources Ascend infrastructure stackDeepSeek published Ascend-optimized versions of its NVIDIA-proven infra components — TileLang Ascend, DeepGEMM-Ascend…infra · JEV 0.65Hindsight: agent memory that learnsAgent memory system built for learning over time: retain/recall/reflect operations with SOTA scores on the LongMemEval…infra · JEV 0.63
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.