Novita AI

Novita AI is a cloud inference platform aimed at developers and startups shipping with open-source models. It offers access to thousands of open-source models (the company cites 10,000+) for LLM chat and completions, embeddings, and image/video/audio generation through OpenAI-compatible APIs, plus GPU instances, bare metal, and sandboxed agent environments for heavier workloads. New models are made available through the API on release day, per the company.

It leans into production agent features: tool calling compatible with OpenAI and Anthropic specs, structured outputs, and prompt caching, and its API works with common coding-agent CLIs. In April 2026 the company cited Artificial Analysis provider benchmarks ranking it first on GPQA Diamond for GPT-OSS 120B; treat that as company-cited rather than independent.

Honest limitations: it is a reseller/abstraction layer over open models, so you inherit the underlying models' licenses and quality quirks, and vendor-cited benchmarks and model counts should be read as marketing. Pricing is usage-based, so heavy workloads need the same cost discipline as any inference API.

Sign up at novita.ai, create an API key, and point any OpenAI-compatible client at the Novita endpoint.