modernbert

Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.

Native MLX runtime for Laya typed decision models — 7–14 ms short decisions on M3 Max. No text generation, PyTorch, or cloud API.

Local Laya typed decisions on Apple Core ML and Neural Engine. Validated ports, ~5 ms short decisions on M3 Max, reproducible speed and energy benchmarks.

Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)

Provenance-first extractive RAG: return verbatim source spans with citations using local ModernBERT or optional LLM-assisted extraction.

Non-autoregressive decision engine on ModernBERT (151M) with calibrated uncertainty (RLCD), TypeSafe AI Jev benchmark audit, and in-browser WebGPU playground

Open reproduction of TypeSafe Jev: a 150M typed decision engine (noul/choice/score in one non-autoregressive pass, calibrated confidence). 0.697 vs Jev's 0.727, 2.5x better calibrated, 4x faster, free. Trains on a Colab T4 in 30 min.

ModernBERT model optimized for Apple Neural Engine.

斑海豹 — Phocinae-Largha-150M-v1: a 144.3M bilingual decision model. Tiny model. Big decisions. 小海豹，大决断。

Score your resume against any number of jobs locally — free, explainable, CPU or GPU. Powered by a 156M ModernBERT cross-encoder.

ClassyText is a demo for zero-shot text classification using ModernBERT-large from Hugging Face.

Train a 1.1M ModernBERT encoder that plays Doom in real time on a CPU

Typed decisions (choice/score/noul) from one encoder pass - faster, better-calibrated successor to Laya, Jev wire compatible

End-to-end pipeline that identifies specialized research papers through automated classification, demonstrated with an LLMOps use case that includes data ingestion, model training, evaluation, and deployment.

Semantic code search using vector embeddings. Search your codebase with natural language queries at the function level.

the research synthethizer outer loop

Code for EXIST 2025 Task 1.1, 1.2 and 1.3

A lightweight CUDA inference library for Laya

A fast second opinion for coding agents: 421M decision model (Laya fine-tune) for destructive commands, success checks, tool choice and prompt injection

Detect duplicate & unused code (py, C, Rust, js/ts) via AST hashing, Jaccard sim, and semantic embeddings (ModernBERT, EmbeddingGemma). w hybrid synthesis