LLM Inference & Serving

Bring Grok to your computer.

The fastest way to build with Gemini.

The generative media platform for developers

Speak it. See it. Share it.

Get up and running with large language models.

AI models and products built for real work

Meet Claude, your thinking partner

The AI community building the future.

A unified interface for LLMs.

A one-stop generative AI development platform.

Build on the Claude Platform

Build real-time, natural conversations with Gemini.

One API to call all LLM APIs

Discover, download, and run local LLMs.

A 128GB unified-memory AI PC platform with 1 PFLOP of FP4 compute

Continual Learning After Deployment

AFM model playground

AI agents for your enterprise

One API for all AI models.

One API. 300+ AI Models. All the power you need.

The easiest way to build and scale generative AI applications with foundation models.

Enterprise platform for customizing and deploying efficient language models.

Open-source RL training with ZoRRo acceleration

AI computing SuperPoD for large-scale AI training and high-concurrency inference

Local AI app and inference engine for agents.

AI-native application development platform

一站式人工智能开发平台

LLM API for Baidu AI Studio

The inference platform for AI products.

The fastest AI inference

Use Claude Code as the unified interface for all your LLMs.

Claude on AWS with native billing, IAM, and Managed Agents

Best Subscription for Open Weight Models.

Build and deploy AI applications on Cloudflare's global network.

Cloudflare Workers AI provider for the AI SDK.

A comprehensive LLM serving solution

The best coding agents run on Baseten

The first AI that can earn its own existence, replicate, and evolve — without needing a human.

Bring PyTorch models to Core AI for on-device execution.

CUDA is NVIDIA's parallel computing platform and programming model.

The Data Intelligence Platform

Turn Deployed AI into Runtime Intelligence

DeepSeek R1 + Claude = DeepClaude

A high-performance expert-parallel communication library.

Clean and Efficient FP8 GEMM Library

Diffusion-based speculative decoding for faster LLM inference

State-of-the-art diffusion models for image, video, and audio generation in PyTorch and FLAX.

A tool for condensed AI/ML model architecture summaries.

Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

DeepSeek 4 Flash and PRO local inference engine for Metal, CUDA and ROCm

Run your own AI cluster at home with everyday devices.

Automatically select the best model for each task.

Fine-tune open source models.

Metadata-driven adaptation of vision foundation models for scientific domains.

The fastest way to build with generative AI

Bringing Flash Magic to Classical Machine Learning Operators

Efficient MLA decoding kernels for Hopper GPUs.

High-throughput tensor loading for PyTorch -- load models at up to 25Gbps without GDS.

High-performance linear attention kernels built on TileLang.

A fully AI-generated pretraining framework.

Pure Rust local OCR for agents

The AI platform for work.

Coding-focused access to GLM models

Run AI models locally on your Android device.

Chat with local LLMs, privately.

Fast AI inference in the cloud

Superpowers for Claude Code

Benchmark LLM inference performance

The platform for scalable, reliable, and efficient inference

A model built for deterministic tasks

The Universal Gateway for AI Inference

The full-stack platform for training your own models

Control every model call

Long Context Pre-Training with Lighthouse Attention

Idea to AI product, ⚡️ fast.

LLM inference in C/C++

LM Studio’s headless daemon

Use your local models, remotely.

Supercharge your LLM with KV Cache.

Multimodal AI software product

Meta Model API public preview

An asynchronous reinforcement-learning framework for large language models

Augment your apps with AI

Build your frontier with Studio.

One AI API, flattened parameters.

ModelScope: discover and use AI models, datasets, and applications.

AI infrastructure that runs everywhere.

A KVCache-centric disaggregated architecture for LLM serving.

Fast inference for coding agents

Ship Apps. Agents. Companies. One prompt. One config file. All your infrastructure.

Scalable reinforcement learning for LLM post-training.

AI Model Aggregation Management and Distribution System

Everything to power your Hermes Agent

Runtime for AI agents to execute real-world tasks

Efficient and accurate low-precision inference.

The AI supercomputer for every enterprise.

Desktop AI supercomputer for AI developers.

AI inference microservices

Free AI Image & Video Studio

A fully open-source framework for instruction tuning and preference tuning of LMs.

Open-source specification and ecosystem for building multi-provider, interoperable LLM interfaces.

Guarantee long-term access to OpenAI compute.

Build with OpenAI

Open-source theoretical reconstruction of the Claude Mythos Recurrent-Depth Transformer architecture

Optimizing LLMs.

One gateway. Every model.

AI Toolkit

Fast speech recognition with NVIDIA's Parakeet models in pure C++.

Enabling robots to understand, remember, and act.

Everything your inference stack is missing.

Machine Learning Platform for AI

A lightweight text-to-speech model.

Open infrastructure for AI research and superintelligence.

PrismML platform

Define functions in English. Run them locally.

Native NVIDIA CUDA kernels for Ampere and newer GPUs.

Build fast, scalable, and cost-effective LLM services.

A coding agent you can leave running.

Advancing biological preparedness with trusted developers and government partners.

The Cloud Built for AI

A fast serving framework for large language models and vision language models.

Inference optimization endpoint targeting lower-cost execution while preserving reference-model behavior.

Edge AI platform

Open Medical Superintelligence

Launch, manage, and stop LLM inference workloads on one or more NVIDIA DGX Spark systems — no Slurm, no Kubernetes, no fuss.

From Coding to Agents, Build It All

Amortized KV Cache Compaction in a Single Forward Pass

The full-stack communications platform.

AI inference at the speed you want and the margin you need

Compute for every scale.

Thinking Machines software platform

Open-source TileLang kernels for DeepSeek workloads

A domain-specific language designed to streamline the development of high-performance GPU kernels for AI workloads.

Ascend TileLang adapter

The AI Acceleration Cloud

Train Dreambooth models. Generate unlimited AI assets.

Automatic model routing for lower LLM spend.

Fine-tune LLMs and VLMs faster with less memory.

Run and train AI models locally with Unsloth Studio.

The world's largest GPU marketplace.

Private, unrestricted access to all the leading AI models across text, image, video, and audio, behind one API key.

A high-throughput and memory-efficient inference and serving engine for LLMs.

Production inference for encoders, poolers, and structured prediction — as vLLM plugins.

High-performance serving framework for omni-modal models.

Fast AI inference for open models

One API for all AI models.

A full-stack platform for open superintelligence.

Serverless inference for frontier open-weight models focused on long horizon agentic workloads.