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.
Speka
Speka uses https://speka.me/v1 with bearer authentication and OpenAI-compatible
chat completions (including streaming) and embeddings. Add a key in the Keys
page or import it as SPEKA_API_KEY in a key file; speka is also recognized
in imported auth JSON. This provider is not Speko.ai.
The Free plan advertises $1 of model usage per month, shared across models, with no card required. It is not unlimited or a separate grant for each model. The pricing page also describes paid overage: check the account's billing settings rather than assuming all requests beyond the grant remain free.
Model rows are published through the signed hosted catalog only. Adding the adapter or importing a key does not seed models or bypass the existing Premium-now / Free-after-30-days release gate. Older app versions must update to recognize the provider. Image generation is not enabled by this adapter.
Key validation uses a nonexistent model against the authenticated chat
endpoint. A controlled 400/model_not_found confirms authentication without
running inference. /models is public and cannot validate a key. Authentication
errors are rejected; quota errors and outages remain inconclusive.
The embedding integration preserves the requested model and checks vector
dimensions through the existing embedding router. The separately tested
nvidia/nemotron-3-embed-1b route returned 2,048-dimensional vectors. Do not
assume interchangeability with another model merely because dimensions match.
Sources, checked 2026-09-24: API reference, pricing.