Skip to content
Docs

Start building with
the Ablitron API.

Ablitron speaks the OpenAI chat-completions format. Point the base URL at Ablitron, add your key, keep the rest of your code.

Quickstart

Connect your tool.

Pick your client, copy the exact config and add your API key. Both models use the same OpenAI-compatible endpoint.

1 · Install the coding client

npm install -g opencode-ai

2 · Save as opencode.json in your project, run opencode, then /connect

{
  "model": "ablitron/ablitron-pro",
  "provider": {
    "ablitron": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Ablitron",
      "options": {
        "baseURL": "https://ablitron.com/v1"
      },
      "models": {
        "ablitron-pro": {
          "name": "Ablitron Pro"
        },
        "ablitron-core": {
          "name": "Ablitron Core"
        }
      }
    }
  }
}
API reference

The details.

Endpoints, limits, billing and error codes for the Ablitron gateway.

Connect a text chat-completions client to Ablitron. Register in the console and fund your balance to use the gateway.

Models

ablitron-pro (1M-token context, based on a modified GLM-5.3) and ablitron-core (256K-token context, half the price) are abliterated models. These are service names, not claims that Ablitron trained the underlying model. Coding quality remains to be benchmarked.

1. Create an API key

Open the console, register and verify your email. Create a key, copy it once and keep it in your server environment. Keys can be revoked from the same panel. Add prepaid credit or choose a monthly subscription.

2. Send a request

Production base URL: https://ablitron.com/v1. For local development use http://localhost:4174/v1. Store your key in ABLITRON_API_KEY.

curl https://ablitron.com/v1/chat/completions \
  -H "Authorization: Bearer $ABLITRON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"ablitron-core","messages":[{"role":"user","content":"Write a Python binary search."}],"max_tokens":2048}'

3. Stream or call tools

Add "stream": true to receive server-sent events with completion chunks, final usage and a [DONE] marker. Function definitions use the tools array; return tool results in messages with role tool and the matching tool_call_id. The client executes tools; Ablitron does not execute your code.

Input and output limits

The preview supports text messages only, at most 10,000 messages and 64 tools. The output limit is 1–16,384 tokens, default 2,048. Use either max_tokens or max_completion_tokens, not both. There is no additional preview token cap on input. The provider enforces the selected model's context window. HTTP request bodies are limited to 16 MB, and a plan with room in its limits, or enough prepaid credit for the request reservation, is required. Only one completion per request is supported.

Plans, credit and billing

With a plan, each request counts toward its 5-hour, weekly and monthly limits, measured at the token rates. When a limit is reached, or without a plan, requests use prepaid credit: they reserve credit before inference, using a conservative input size and the maximum requested output, and the final charge uses reported token usage. Cached input is a subset of input, charged at its lower rate. Reasoning tokens are already included in output and are not charged twice.

Errors and retries

400
Invalid or unsupported request. Check the message and output limits.
401
Missing, revoked or inactive key.
403
account_suspended: the account behind the key is suspended. Contact support.
402
Insufficient available credit for the reservation. Add credit or lower the output limit.
429
Rate limit (120 requests a minute per network and per key by default, or the key's own key_rate_limit), four requests already in flight, key_usage_limit: the key reached the daily or monthly usage limit set for it in Settings → API keys, or usage_limit: your plan's limit is reached and you have no prepaid credit. error.resets_at (Unix time) and the Retry-After header say when it renews.
502
Model response or usage could not be confirmed. Check your console before retrying.
503
Inference or another required service is not enabled.

Save the X-Request-ID response header when reporting an issue. Inference is not automatically retried. If usage cannot be confirmed after a disconnect or server error, the reservation stays pending for 15 minutes; it is then settled at a conservative estimate of the text you received, or released without charge if nothing was delivered. Requests without streaming end after about 170 seconds with 504 request_timeout; the prompt is charged at an estimate, so use streaming for long answers. A retry is a new request and may incur additional usage.

Data handled by the gateway

Requests are forwarded to an inference provider. The local usage ledger stores request IDs, timestamps, model aliases, token counts and charges; it does not store prompt or output text. Customer API keys are stored as hashes. Provider-side handling and applicable service terms must be reviewed for your use case; no end-to-end zero-retention claim is made.

Ask anything.
Get an answer.

Start chatting in seconds. No card required.