40 providers list GPT-5.6 Luna — none is a verified unlimited free API
Asked:
Omg use Provider with ai Gateway dont use Codex oder find unkown Provider find 40 api Provider and coding
40 API providers, gateways, routers, and clouds listing GPT-5.6 Luna access were reviewed as of 2026-09-03. Verdict: 0 fully verified unlimited free APIs, 1 unverified free candidate (Kenari), 11 confirmed paid, 1 paid promotion, and 27 with no proof of ongoing free use. Codex is excluded as requested.
Verdict across all 40 providers
Kenari — verify before use. One independent pricing tracker marks this gateway “Free”, but its own terms, limits, signup requirements, and current endpoint could not be independently verified. Treat it as a free candidate only, never a confirmed unlimited service — llmpricing.dev
Findings
Zero of the 40 reviewed providers could be verified as an unlimited, ongoing, no-payment API for GPT-5.6 Luna —
apirank.vip
Free-sounding names are not free: Free.ai states token-based pricing, FreeModel has no verified ongoing-free terms, and OrcaRouter explicitly says no free tier —
free.ai ·
techbullion.com
Cheapest confirmed price is NanoGPT at $0.10 in / $0.60 out per 1M tokens; six gateways including OpenRouter and Cloudflare AI Gateway list $0.20 / $1.20 —
llmpricing.dev
Infrastructure is not an IDE: an OpenAI-compatible endpoint (Venice AI, OneMux and others) can usually be wired into coding tools that accept a custom base URL and API key, but few provider pages confirm a specific IDE —
onemux.net
Exact model identifiers seen across providers
gpt-5.6-lunaopenai/gpt-5.6-lunaopenai.gpt-5.6-lunaopenai-gpt-56-luna
Safety before testing any “free” provider
- Never enter payment details or valuable credentials merely to test an unverified provider.
- Create a fresh low-privilege API key, and set hard spend limits before the first call.
- Run a tiny test request first and check the bill and the response before any real use.
Final review dataset: 40 rows, one provider per row, compiled 2026-09-03. Each row records the listed GPT-5.6 Luna model ID, access type, cost verdict, coding readiness, restrictions, and count of supporting pages found. Verdicts reflect what public pages state; no endpoint was live-tested. Coding-readiness prose is shortened in the table for space. Codex is excluded by request.