Aider

Checked against Aider's docs on 2026-10-11

Aider connects as an OpenAI-compatible endpoint: two environment variables and the model name with Aider's openai/ prefix.

Base URL: https://api.creeksidecompute.ai/v1
API: OpenAI Chat Completions
Model id: qwen/qwen3.8-27b, written openai/qwen/qwen3.8-27b in Aider, or openai/qwen/qwen3.6-35b-a3b for speed, or openai/meta/muse-glimmer-30b, which always reasons first
Context window: 131072 tokens, up to 65536 of them output
Key: Make one at creeksidecompute.ai/account

Set it up

Make a key. Sign in at creeksidecompute.ai/account with your email and make a key. Each request is paid from prepaid credit: top up by card from $10, or redeem a promotional code.
Set the variables. Put them in your shell profile and open a new terminal, or put the same names in a .env file in your home folder or your repository:
shell
```
export OPENAI_API_BASE=https://api.creeksidecompute.ai/v1
export OPENAI_API_KEY=<your key>
```
Aider sends every OpenAI model to this base URL while these are set. If OPENAI_BASE_URL is also set, it wins: remove it or give it the same value.

Tell Aider the model's limits. Put this in .aider.model.metadata.json in your home folder or your repository. Without it, Aider warns that it does not know the model:

.aider.model.metadata.json

{
  "openai/qwen/qwen3.8-27b": {
    "max_tokens": 65536,
    "max_input_tokens": 131072,
    "max_output_tokens": 65536,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/qwen/qwen3.6-35b-a3b": {
    "max_tokens": 65536,
    "max_input_tokens": 131072,
    "max_output_tokens": 65536,
    "litellm_provider": "openai",
    "mode": "chat"
  },
  "openai/meta/muse-glimmer-30b": {
    "max_tokens": 65536,
    "max_input_tokens": 131072,
    "max_output_tokens": 65536,
    "litellm_provider": "openai",
    "mode": "chat"
  }
}

Run Aider in your repository: aider --model openai/qwen/qwen3.8-27b. The openai/ prefix tells Aider which kind of endpoint this is; Aider removes it before it sends the model id. In a session, switch with /model openai/qwen/qwen3.8-27b. For speed, use openai/qwen/qwen3.6-35b-a3b instead, and for Muse Glimmer 30B, openai/meta/muse-glimmer-30b.

Thinking

Thinking is off unless Aider asks for it. To let --reasoning-effort through for this model, put this in .aider.model.settings.yml beside the metadata file:

.aider.model.settings.yml

- name: openai/qwen/qwen3.8-27b
  accepts_settings: ["reasoning_effort"]
- name: openai/qwen/qwen3.6-35b-a3b
  accepts_settings: ["reasoning_effort"]
- name: openai/meta/muse-glimmer-30b
  accepts_settings: ["reasoning_effort"]

Then aider --model openai/qwen/qwen3.8-27b --reasoning-effort medium turns thinking on, and none keeps it off. In a session, /reasoning-effort high or /reasoning-effort none switches it. Aider uses the same model for commit messages and summaries, so those think too. Thinking tokens are billed as output. Do not use --thinking-tokens: it sends Anthropic's thinking setting, not the switch this endpoint reads.

Muse Glimmer 30B always reasons before it answers, so --reasoning-effort sets how long, not whether: without the flag, or at none or low, it reasons briefly, and medium and high make it reason longer.

What does not work

Images, on the Qwen models. They are listed for text only.
Streamed thinking. While Aider streams, it may not show the thinking text; --no-stream shows it above the answer.
Cost display. Without prices in the metadata file, Aider shows token counts and no cost. Ours are in GET /v1/models.

Troubleshooting

401 invalid API key: The key is wrong, or Aider did not get it. Check OPENAI_API_KEY in the shell or .env file that Aider reads; a .env value wins over the shell. To test the key alone: curl -s https://api.creeksidecompute.ai/v1/models -H "Authorization: Bearer $OPENAI_API_KEY".
404 model … is not served: The model id is wrong. Use qwen/qwen3.8-27b exactly, in lower case, with the qwen/ prefix. GET /v1/models lists every id we serve. In Aider the name is openai/qwen/qwen3.8-27b: the prefix, then the id.
400 maximum context length is 131072 tokens: The conversation is longer than the model's window. Keep max_input_tokens at 131072 in the metadata file so Aider warns before a request is too big, then drop files with /drop or clear the chat with /clear.
402 or 429: A 402 means your prepaid credit does not cover the request: add credit at creeksidecompute.ai/account. A 429 means the key's rate, or the number of requests your account may run at once, is used up; agents that work in parallel reach that sooner. Wait and retry. The API reference lists the limits.

Other tools

Sources, read on 2026-10-11: OpenAI compatible APIs, advanced model settings, reasoning settings.