This is the editorial ranking: not an automatic filter, but the twenty free tiers we would actually reach for first. Each pick links to its full verified profile — free tier, rate limits and the catch are all checked against the provider's own docs on the linked page. The list is reviewed and re-verified with the dataset.
Typhoon (SCB 10X)
Editor's pickThe rare package that is genuinely free forever (perpetual, not a trial) with a frontier-class model, OCR and real-time speech behind one no-card OpenAI-compatible key.
Cloudflare Workers AI
Best free quotaThe best free daily volume for a real side project — 10k requests a day with no card, and the quota renews.
Google Gemini API (AI Studio)
Frontier qualityThe only frontier-class model with a genuine, renewing free tier — the default when you want the smartest free model.
Ollama Cloud
Best on-rampCloud-hosted open models with the exact Ollama workflow and no card at signup — the easiest on-ramp for open-weight models.
OpenRouter
Best gatewayOne no-card key, every model: a wide rotating sample of :free models behind a single OpenAI-compatible endpoint.
AI Horde
No account neededAnonymous, no-signup and permanently free — the fallback that works when everything else asks for an account.
LlamaParse (LlamaCloud)
Best for RAGThe most generous free document-parsing tier for RAG builders (~10k pages a month), renewing every month.
OCR.space
Reliable utilityClassic, reliable free OCR at 25k conversions a month with no card — the boring tool that just works.
Speechify API
Best for TTSFree commercial TTS at 50k characters a month — the pick when the output actually ships.
Jina AI
Best for embeddingsFree embeddings and rerank behind an OpenAI-compatible endpoint — the shortest path to a free RAG stack.
Deepgram
Best STT creditThe largest speech credit in the list — $200 for Nova STT and Aura TTS with no card and no expiration, plus an opt-out from model training.
Groq
Fastest inferencePurpose-built LPU inference makes this the lowest-latency free tier for open-weight models — the only catch is phone verification at signup.
SambaNova Cloud
Fast, no cardA rate-limited free tier across all models with a commercial license and no card — fast open-model inference without a trial clock counting down.
W&B Inference
Best monthly frontier credits$100/month of serverless inference credits covering frontier open models behind one OpenAI-compatible key, with Weave tracing on top.
Voyage AI
Best embedding allotment200M one-time tokens across best-in-class embedding and rerank models — the biggest free embedding balance here, no card required.
Pinecone Inference
Best managed embeddings5M embedding tokens and 500 rerank requests a month on a managed Starter plan — a real, renewing free tier for RAG.
Pollinations.ai
Free images, no signupHosted Flux and Stable Diffusion behind a simple GET URL, anonymous to start — the easiest way to generate free images programmatically.
Moondream Cloud
Best tiny visionA $5/month recurring credit for the Moondream vision model — caption, VQA and detection behind an OpenAI-compatible endpoint.
Unstructured
Best doc pipeline15,000 pages a month of document parsing and OCR with no card — the free path from messy PDFs to LLM-ready structured data.
Scaleway Generative APIs
Best EU open models1M free tokens and 60 minutes of Whisper on EU-hosted open models — the sovereignty-friendly free tier for European workloads.
Editorial ranking, reviewed 2026-08-14 — every pick links to its full verified profile. Want to nominate a provider? See the repo.