{
  "id": 8955640,
  "title": "Free LLM gateways, tested: what the free tier actually gives you in September 2026",
  "url": "https://urgent.news/2026/09/21/free-llm-gateways-tested-what-the-free-tier-actually-gives-you-in",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-21T16:08:16.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/yummy342/free-llm-gateways-tested-what-the-free-tier-actually-gives-you-in-september-2026-1l50"
  },
  "original_language": "en",
  "account": "In a September 2026 analysis of six free LLM API gateways, it was found that each provider's free tier offers different limitations and features. OpenRouter's free tier starts at 50 requests per day, increasing to 1,000 after $10 in lifetime credit. Cloudflare AI Gateway provides no analytics, caching, rate limiting or retries for free. Vercel AI Gateway offers a monthly free credit, but the amount is undisclosed. LiteLLM is a free software requiring users to supply provider keys, server, database, and maintenance. FreeModel by Aiglade has no free capacity; it shifts to paid fallback after running out. Every link in the article directs to the source page stating the numbers, allowing verification. OpenRouter's free tier varies based on account credit, making the 50 requests/day figure irrelevant for growing needs. LiteLLM's model count is not comparable across products. The list is dynamic, with models being added, retired, and quotas changing. The actual free tier effectiveness depends on the endpoint's honesty, the fallback mechanism, and the user's infrastructure. FreeModel's key offering provides free capacity with paid fallback, along with OpenAI and Anthropic compatibility. The comparison article also discusses when each gateway might be the wrong choice.",
  "summary": "Free LLM gateways, tested: what the free tier actually gives you in September 2026 Every \"free LLM API\" roundup I can find has the same two problems. It quotes limits from the provider's marketing page instead of the API's own responses, and it treats the model count as a number you can compare across products. Both break within weeks. So I checked the endpoints directly. Here is what six…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Dev.to",
        "title": "What AI engines actually cite when you ask about LLM gateways",
        "url": "https://urgent.news/2026/09/21/what-ai-engines-actually-cite-when-you-ask-about-llm-gateways",
        "published": "2026-09-21T16:07:12.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}