Free LLM APIs Compared: Rate Limits, Models, and Real Costs (2026)

Compares 13 free LLM API platforms including OpenRouter, Google AI Studio, Groq, and Mistral, examining rate limits, models, context windows, and hidden costs like data-training policies. Highlights Groq’s LPU hardware running Llama 3.3 70B at around 320 tokens per second, Google AI Studio’s up to 1 million token context, and Mistral’s roughly 1 billion free tokens monthly with data-training opt-in; OpenRouter offers one OpenAI-compatible endpoint for 20+ free models.