freellmapi on Anyken
freellmapi on Anyken: compare the model examples with your intended input and output. Review the capabilities documented below, then use the controls actually offered in the workspace. Available settings and credits are shown before submission.
Read source
freellmapi: Developer Overview
This section covers the core architecture and value proposition of freellmapi for developers who need to aggregate free-tier LLM capacity without managing multiple provider integrations.
Aggregated Free-Tier Capacity Behind One Endpoint
This project collapses 34 free LLM providers and 635 free model endpoints into a single local /v1 API. The combined theoretical free-tier capacity across all supported providers is roughly 7.4 billion tokens per month, though your actual total depends on which accounts you configure. Each provider on its own offers a small quota—a few million tokens a month, a few thousand requests a day—but stacked together they become a meaningful prototyping resource. The router tracks per-key RPM, RPD, TPM, and TPD counters that learn each provider's reported ceilings, so routing always stays under every cap. When a provider returns a 429 or 5xx, the router cools that key down and immediately retries the next model in your fallback chain. Provider keys are encrypted at rest using AES-256-GCM in a local SQLite database and decrypted in memory only for the duration of a request. Your applications never see the upstream keys—they only use a single unified bearer token generated by the dashboard.
Self-Updating Model Catalog and Routing Strategies
The router pulls a signed model catalog from freellmapi.co twice a day, applying new models, quota changes, and provider quirk fixes automatically. Free installs receive the monthly snapshot, meaning a model reaches them 30 days after it lands in the live feed; premium routers ($19/year) get same-day updates. The catalog currently tracks 474 model families across 34 providers, including 584 chat endpoints, 41 embeddings endpoints, 7 transcription endpoints, and 3 video endpoints. Six routing strategies are available: Manual, Balanced, Smartest, Fastest, Most Reliable, and Custom. The Balanced strategy (recommended default) weights reliability at 50%, speed at 25%, and intelligence at 25%. The adaptive router uses Thompson Sampling, a bandit method that balances exploiting well-performing routes with exploring alternatives that may now be better. Each response carries an X-Routed-Via header showing which provider and model actually served the request, giving full observability into routing decisions.