Open-source framework for training harness-based agents end-to-end in diverse environments.

A family of native computer use agents (CUAs) for scalable browser and computer interaction.

A framework for programmatically optimizing prompts and weights in LLM pipelines.

Multimodal frontier model for synchronized video, audio, and image generation with robotic action prediction.

Extensive tutorial resource for learning Modern OpenGL

A step-by-step guide to learning the Vulkan graphics API

A trainable agentic framework that coordinates planner, executor, verifier, and generator modules.

A one-step diffusion model framework for efficient, high-quality text-guided video editing.

Scaling Latent Reasoning via a Surrogate Policy for autoregressive latent reasoners.

A two-layer memory architecture combining profile expansion and compression residuals for LLM agents.

Training-free sub-quadratic prompt compression via Krylov-projected LSH.

Zero encoding extreme performance binary serializer for C#

Distributed swarm training framework for Agentic RL supporting heterogeneous multi-model training.

A high-throughput, scalable suite of GPU-accelerated assistive robotics tasks built on JAX and MuJoCo-MJX.

Full harness, corpus generator, and raw results for evaluating LLM prompt design at scale.

The largest scientific code corpus and executable benchmark for scientific code generation.

A library for secure multi-party computation to enable private data analysis.

Implementation of compound sparsity framework for LLM compression including pruning and dynamic skipping.

An open-weight 118B MoE coding model with high efficiency and 1M context window.

Hypothesis-driven Agentic Retrieval and Probing for training-free LLM interpretability.