Skip to content

7.4 billion tokens per month. 34 free LLM providers. 635 free model endpoints. All behind one /v1 endpoint, plus any custom OpenAI-compatible endpoint. Smart routing, automatic failover, encrypted keys. Personal experimentation only.

TypeScript
2.3 KB
Raw Blame

Gizmo and BlockRun

Both providers use bearer API keys. Add a key in Keys, or import GIZMO_API_KEY=... / BLOCKRUN_API_KEY=.... Model rows come exclusively from the signed hosted catalog; no model seeds or automatic roster imports ship with these adapters. Existing Premium-now / Free-after-30-days behavior is unchanged.

Gizmo

  • Base URL: https://api.gizmoplatforms.com/api/v1
  • Authentication check: GET /models (authenticated).
  • Free plan: 1,000 zero-cost-model requests per calendar month, shared across account keys and models. Only tested :free routes qualify without a card.
  • Text chat and SSE streaming are supported. Tool calling, image/audio messages, structured-output controls and unlisted sampling fields are not. Non-text/tool messages fail explicitly instead of being silently discarded.
  • The documented default output allowance is 1,024 tokens; upstream maximum is 32,768. Existing operator/context caps still apply.

The six catalog candidates tested on 2026-10-03 are Qwen3.8 27B, Nemotron 3.5 Lightning, Nemotron 3 Ultra, Nemotron 3 Super, Laguna XS2.1 and Laguna S2.1. Successful responses reported zero cost. Metered GLM and DeepSeek routes required a payment method and are excluded.

Official API documentation.

BlockRun

  • API-key base URL: https://api.blockrun.ai/v1 (not the wallet/x402 host).
  • Authentication check: GET /account, without generating tokens. A depleted prepaid balance does not itself invalidate a key or prevent $0 routes.
  • Free routes are ongoing zero-priced, rate-limited models, not monthly cash credits. Paid models in the same roster are not automatically added.
  • The provider can silently substitute a different free model with HTTP 200. Non-streaming responses and every streamed chunk must identify the requested model exactly; otherwise the adapter raises 502 before forwarding that data.

The four exact-model chat candidates tested on 2026-10-03 are Nemotron 3 Nano Omni, Llama 3.2 11B Vision, Muse Glimmer 30B and Laguna XS2.1. Ledger charges settled at $0. Tools/vision capabilities are not inferred from model names; the initial catalog publishes conservatively tested text-chat routes.

API-key documentation and free-model policy.