Aider works with any OpenAI-compatible API. Point it at InferAll and it can use our free models ($0 input, $0 output).
Setup
Get a key at inferall.ai/keys, then:
export OPENAI_API_BASE=https://api.inferall.ai/v1
export OPENAI_API_KEY=ifu_your_key_here
aider --model openai/nvidia/nemotron-3-ultra-550b-a55b
The openai/ prefix tells Aider to use an OpenAI-compatible endpoint. Aider removes it before sending, so InferAll receives nvidia/nemotron-3-ultra-550b-a55b. To use gpt-oss-20b instead:
aider --model openai/openai/gpt-oss-20b
Aider may warn that it does not know the model's context size. That is only a warning; add --no-show-model-warnings to hide it.
What we tested (2026-10-21, Aider 0.86.2)
We ran Aider in a small project and asked: "Add a function farewell(name) to app.py that returns f'bye {name}'. Change nothing else."
| Model | Result | Time |
|---|---|---|
nvidia/nemotron-3-ultra-550b-a55b |
edit applied correctly | 29s |
openai/gpt-oss-20b |
edit applied correctly | 5s |
In the first run Nemotron 3 Ultra was slow to start, so InferAll answered from gpt-oss-20b, another free model, and said so in the response. Times vary with load on the shared free models.
Paid models
With a funded key, the same setup reaches paid models by their prefixed ids, for example:
aider --model openai/anthropic/claude-sonnet-4-6
InferAll is an LLM aggregator: one API key for free open models and paid OpenAI, Anthropic, Google and xAI models. New accounts get 25 free calls on open models, no card needed.