Skip to content

[Provider]: Cheaper Inference #3884

Description

@aiapienthusiast

Provider Name

Cheaper Inference

Provider Website

https://cheaperinference.com

API Documentation

https://cheaperinference.com/docs

API Compatibility

OpenAI-compatible

Available Models

The gateway serves 60 chat routes under their makers' own names, among them:

  • claude-opus-5, claude-sonnet-5, claude-haiku-4.5 (Anthropic)
  • gpt-6-astra, gpt-5.5, gpt-5.6-luna / -sol / -terra, gpt-5-mini (OpenAI)
  • gemini-3.7-flash, gemini-3.1-pro-preview (Google)
  • grok-4.5 (xAI), kimi-k3 (Moonshot), glm-5.3 (Z.ai)
  • deepseek-v4-pro, deepseek-v4.1-flash (DeepSeek)
  • qwen-3-8-max and other Qwen routes (Alibaba)
    Plus 5 image models and 1 video model on their own endpoints.

Key Features

  • one endpoint and one key for models from several makers
  • the model id is the maker's own, with no vendor prefix
  • an authenticated GET /v1/models publishes per-model prices, context and output limits, and capability flags, so a client can read the catalogue rather than hard-code it
  • four models are priced in two bands with a published long-context threshold
  • the catalogue moves: four of the 60 routes changed price in the 23 hours between two readings, which is why reading it beats pinning it

Authentication Method

API Key

Pricing Information

https://cheaperinference.com/docs
Per-model rates come from the provider's own GET /v1/models response, which reports both the charged rate and the maker's list price.

Community Interest

Medium - Some users would benefit

Your Use Case

I use this gateway already and want Forge pointed at it without a custom OpenAI endpoint. What I get from it is one key for models from several makers, which matters when a session moves between a cheap model for bulk edits and a frontier one for the hard part - today that means re-configuring the provider mid-session. The models are addressed by their makers' own names, so the same model id works whether I go through the gateway or direct.

Contribution

  • I'm willing to submit a PR to implement this integration
  • I can help with testing
  • I have access to this provider's API for testing

Additional Context

The change follows #3623 (OrcaRouter): an entry in crates/forge_repo/src/provider/provider.json, the ProviderId in crates/forge_domain/src/provider.rs, the repo wiring, and a README row. Models are discovered from the provider's own /models endpoint, so no model list is hard-coded. 618 + 334 tests and clippy are green locally.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    type: featureBrand new functionality, features, pages, workflows, endpoints, etc.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions