GitHubGitHub RepoNeeds Review

tashfeenahmed/freellmapi

FreeLLMAPI self-hosts one API over many providers’ free tiers, with routing, failover, quota tracking, and coding-agent setup. Its capacity figures are project-maintained claims.

GitHub

Open Source Open Markdown Open JSON

Source Summary

FreeLLMAPI claims a catalog of **34 providers**, **635 free endpoints**, and roughly **7.4B listed tokens per month** behind OpenAI-, Anthropic-, Gemini-, and Ollama-compatible interfaces. It adds routing, failover, quota tracking, encrypted local key storage, and agent setup generators.

Practical Implication

For low-cost experiments, point a non-critical Codex or Claude Code setup at the local router and inspect the routed-provider header. Test tool calls, structured output, model switching, and rate-limit behavior before trusting a fallback chain with repository work.

Agent-Ready Context

FreeLLMAPI claims a catalog of **34 providers**, **635 free endpoints**, and roughly **7.4B listed tokens per month** behind OpenAI-, Anthropic-, Gemini-, and Ollama-compatible interfaces. It adds routing, failover, quota tracking, encrypted local key storage, and agent setup generators.

For low-cost experiments, point a non-critical Codex or Claude Code setup at the local router and inspect the routed-provider header. Test tool calls, structured output, model switching, and rate-limit behavior before trusting a fallback chain with repository work.

The capacity is aggregated from documented free tiers, not a guaranteed pool, and provider quotas change frequently. Free installs receive catalog additions after a **30-day delay**, while the live feed is paid; upstream providers still receive requests sent through them.

Connected Context · Feed7 Judgment

FreeLLMAPI broadens low-cost model experimentation by aggregating many changing free tiers behind familiar APIs, but it is a local routing catalog rather than guaranteed capacity or a privacy boundary. Against managed gateways, its differentiator is breadth and client compatibility; its tradeoffs are delayed free catalog updates, upstream data exposure, volatile quotas, and the need to validate fallback behavior per workload.

MiniMax M3 and M2.7 are free on AI GatewayThe expiring MiniMax routes reinforce that zero-cost endpoints are unsuitable as assumed durable capacity; FreeLLMAPI fallback chains likewise need explicit handling for quota or catalog changes.AI Gateway adds unified fast mode supportBoth make routing metadata operationally important: Vercel clients must confirm whether fast mode was used, while FreeLLMAPI users should inspect which upstream provider actually served a request.Kimi K3 and Kimi K3 Fast with ZDR and US-based providers now on AI GatewayKimi’s explicit US-hosting and ZDR options contrast with FreeLLMAPI’s model, where requests still reach upstream providers; privacy and residency therefore require provider-level evaluation.Gemini 3.7 Flash now available on AI Gateway for 50% offThe temporary Gemini discount provides a managed low-cost comparison with routing, retries, and spend controls, while FreeLLMAPI offers broader free-tier aggregation whose quotas are not guaranteed.

Context Map

infracoding#gateways#coding-agents#model-selection

Uncertainty

The capacity is aggregated from documented free tiers, not a guaranteed pool, and provider quotas change frequently. Free installs receive catalog additions after a **30-day delay**, while the live feed is paid; upstream providers still receive requests sent through them.