Free image, speech & embedding APIs

The 5 fastest free LLM APIs (no card, no KYC)

You want a working chat or completions call in under two minutes. Five providers score 100 on our Ease scale: email signup, instant key, zero billing. Here they are, with the real free-tier numbers we verified. For the broader landscape, our Best Free AI APIs guide ranks providers by use case, and How to pick a free LLM API walks the tradeoffs.

• Groq — 30–60 RPM and 14,400 requests/day on LPU hardware. No card. The fastest raw free tier in the table.
• Google AI Studio — 5–30 RPM / 1,500 requests/day on Gemini, 1M–2M context. Requires a Google account (new accounts need a phone).
• OpenRouter — one key, 25+ free :free models behind an OpenAI-compatible endpoint. 20 RPM / 50 RPD.
• Cerebras — ~1M tokens/day (about $15 of inference) on Llama and Qwen. No card.
• Hugging Face Inference — 100+ open models, $0.10/mo routing, 300 requests/hour serverless.

OpenRouter alternatives: one key, many models

OpenRouter is the popular gateway, but it is not the only one. If it is rate-limited or down, these drop-in replacements give you many models behind a single API key, and several run their own free tiers:

• OpenRouter — the baseline: 25+ free models, 1K RPD after a $10 top-up.
• LiteLLM — self-hosted proxy for 100+ LLMs, unlimited, no free-tier cap (you pay the upstream provider).
• Pollinations — unlimited, no auth, text and image from a URL.
• OrcaRouter, bazaarlink, 9Router, Portkey AI, OmniRoute — gateways with 40–250 providers and free tiers.

Pick a gateway when you want to swap models without rewiring your code. Pick a direct provider (Groq, Cerebras) when you want one model at maximum speed.

Free image, speech, and embedding APIs

The table is not just LLMs. The same Ease score applies to the rest of the AI stack:

• Deepgram — $200 credit for speech-to-text.
• AssemblyAI — $50 credit, 5 concurrent streams.
• ElevenLabs — 10,000 characters/month on the free plan.
• Embeddings: Voyage AI (50M tokens), Jina (1M tokens), Pinecone and Weaviate (free vector stores).
• Images: Recraft, Freepik, Leonardo AI (150 tokens/day), Ideogram all run free tiers.

Rate limits change — how we keep this honest

Free tiers move. A provider bumps its RPM, adds a card gate, or retires a model, and this table goes stale. We re-verify on a weekly pass and mark status as active, gated, trial, beta, or retired. The raw data is published as JSON at /data/llm-providers.json so you can build your own monitor. Always confirm the current limit on the provider's pricing page before you ship a product on a free tier.