Agent infrastructure: vector DBs, inference, observability, and search.

Open-source machine-learning framework

The easiest way to run open language models locally

Reference vector database for RAG and semantic search — Starter tier is free up to 2GB

The model hub the open-source AI ecosystem runs on — free Spaces, $9 PRO, $20/user Team

Simba TTS API at 300ms latency and $10 per 1M chars

Type-safe Python agent framework

Run, fine-tune, and deploy AI models with one line of code

Fast, low-cost inference for open-source models

Sub-second LPU inference — Llama 3.1 8B at 840 tokens/sec for $0.05/M input

Open-source framework for building LLM applications

Serverless GPU compute for AI builders

Meta's open-weights family of foundation models

Open-source LLM observability — 10K free requests, OpenAI/Anthropic/Together drop-in proxy

Open-source LLM observability and evals — Hobby tier free, $29/mo Core, self-hostable

Web search API designed for AI agents — 1,000 free credits/mo, $0.008/credit PAYG

Cheap, fast inference for open models — Llama 3.3 70B at $0.88 per million tokens

Google's open-weights model family

Alibaba's open-weights language model family

Experimental self-building autonomous agent framework

Sub-quadratic LLM for 12M-token reasoning