How do I configure Codex CLI with FreeModel? Two environment variables

Setup

# OpenAI-compatible base. Note: /v1 is part of the base —
# the client appends /chat/completions itself.
export OPENAI_BASE_URL="https://freemodel.online/v1"
export OPENAI_API_KEY="sk-your-key"

Restart the shell that launches Codex CLI, or put the two lines in your shell profile so they survive a new terminal.

Which model you get

Whatever you ask for. If you name a specific model, that is what answers. If you would rather not track model names, pass a tier alias instead:

The full set of ids is at Models, and the live list is curl https://freemodel.online/v1/models.

Checking it works

# does the base answer at all? 401 = yes, it wants a key. 404 = wrong path.
curl -s -o /dev/null -w "%{http_code}\n" \
  -X POST https://freemodel.online/v1/chat/completions \
  -H "Content-Type: application/json" -d '{}'

# and a real call, which proves the key
curl -s https://freemodel.online/v1/chat/completions \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{"model":"fm/qwen-plus","messages":[{"role":"user","content":"hi"}]}'

A 400 with messages: ... required on the first command is the good outcome — it means the endpoint exists and parsed your (empty) body.

Common problems

Next

Set up in three steps

The whole configuration is two environment variables. The order matters only in that the shell which exports them has to be the shell that launches the client.

1. Step 1: Export the base and the key. OPENAI_BASE_URL takes https://freemodel.online/v1 and OPENAI_API_KEY takes your sk- key from the console. The base stops at /v1; the client appends /chat/completions on its own.
2. Step 2: Make it survive a new terminal. On macOS and Linux that means the same two lines in ~/.zshrc or ~/.bashrc. On Windows, setx OPENAI_BASE_URL "..." writes the value for future shells, while $env:OPENAI_BASE_URL covers only the current PowerShell session.
3. Step 3: Restart Codex CLI and check the endpoint before you blame the model. curl -s https://freemodel.online/v1/models is public and needs no key, so it separates a reachability problem from a key problem in one command.

Codex CLI is a coding agent, so whatever answers has to hold up across a long run. The 199 listed models are 77% chat models[1], and a routed request is re-decided on every call, so a provider that stops answering mid-session is skipped rather than fatal.

What you get to choose from

Naming a model pins the request to it. Naming a tier alias hands the decision to the router per request, and the aliases are a deliberately small slice of the catalogue — three of them, one per subscription plan.[1] That is the point of them: they stand in for a choice you would otherwise make yourself, and they move when a provider stops answering.

The wider catalogue is worth a look if you care about depth rather than convenience. The 199 listed models put chat, image, video, speech and embeddings in one list, and six auto/* routes cover the non-chat side — auto/image, auto/video, auto/tts, auto/asr, auto/embed, auto/rerank — each carrying a candidates field that shows how many models sit behind it.[1]

A route is not a model name: it is a standing instruction to the router, re-evaluated on every call. Two identical requests inside one session can be served by two different providers, which is what you want for uptime and not what you want while comparing two prompts against each other.

Codex CLI FAQ

Sources

Every third-party figure quoted on this site was read from the vendor documentation linked below, on the date stated. None of these links are affiliate links and none of these vendors paid to be listed. If a number here disagrees with the page it came from, the page is right — tell us and it gets corrected.

• OpenRouter — credit-purchase fees and free-model rate limits
• OpenRouter — rate-limit reference
• OpenRouter — public model list
• LiteLLM — licence, deployment options and supported models
• Cloudflare AI Gateway — features and pricing
• Vercel AI Gateway — features
• Portkey — plans
• Helicone — plans
• Model Context Protocol — specification
• OpenAI Python SDK — base_url and OPENAI_BASE_URL
• Anthropic Messages API — request format

Where the percentages come from

Every share quoted on this site is counted from the live model list, not estimated. Divide the two counts in the last column, round to the nearest whole percent, and you get the number in the first.

[1] Counted on 29 September 2026 from GET https://freemodel.online/v1/models, which needs no key and no account. Every row uses all 199 listed entries as the denominator, because the only thing the endpoint states for every entry is its modality. It no longer publishes per-model capability flags such as tool calling or reasoning, so this site quotes no shares for them. Fetch the list and recount whenever you want to check the arithmetic.