Endpoints

Chat completions

Send chat messages to any text model with POST /chat/completions, using the OpenAI chat format, roles, parameters and response fields.

Try it in the playground

The chat completions endpoint uses the OpenAI chat format and works with every text model in the catalog: GPT, Claude, Gemini, Llama, DeepSeek, Mistral and hundreds more. The request body goes to the upstream provider unchanged. api14 only checks your key, model, limits and balance, then records usage when the request finishes.

POST https://api14.com/api/v1/chat/completions
Authorization: Bearer sk-api14-...
Content-Type: application/json

Request

Parameter	Type	Description
`model`	string, required	Model id, e.g. `anthropic/claude-sonnet-4.5`. Variants like `:online` or `:nitro` are allowed. See Model ids.
`messages`	array, required	Conversation as `{ role, content }` objects. Roles: `system`, `developer`, `user`, `assistant`, `tool`.
`models`	string[]	Fallback models to try in order if the first one fails. See Provider routing.
`stream`	boolean	Stream tokens as server-sent events. See Streaming.
`max_tokens` / `max_completion_tokens`	integer	Upper bound on generated tokens.
`temperature`	number	Sampling temperature, usually `0`–`2`.
`top_p`, `top_k`, `min_p`, `top_a`	number	Other sampling controls. Models that don't support them ignore them.
`frequency_penalty`, `presence_penalty`, `repetition_penalty`	number	Repetition controls.
`stop`	string \| string[]	Stop sequences.
`seed`	integer	Best-effort deterministic sampling.
`logprobs`, `top_logprobs`	boolean, integer	Return token log probabilities when the model supports them.
`tools`, `tool_choice`, `parallel_tool_calls`		Function calling. See Tool calling.
`response_format`	object	JSON mode or JSON Schema output. See Structured outputs.
`reasoning`, `include_reasoning`	object, boolean	Thinking controls. See Reasoning.
`plugins`, `web_search_options`		Web search. See Web search.
`provider`	object	Upstream provider preferences. See Provider routing.
`modalities`	string[]	e.g. `["image", "text"]` for image output, or `["text", "audio"]` for audio output.
`prediction`	object	Predicted output (`{ type: "content", content }`) to cut latency on edits.
`transforms`	string[]	e.g. `["middle-out"]` to compress prompts that exceed the context window.
`user`	string	Your own end-user identifier.

Message content can be a string or an array of content parts (text, image_url, file, input_audio). See Vision & files.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api14.com/api/v1",
  apiKey: process.env.API14_API_KEY,
});

const result = await client.chat.completions.create({
  "model": "openai/gpt-5-mini",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

Response

{
  "id": "gen-1760000000-abc123",
  "object": "chat.completion",
  "created": 1760000000,
  "model": "anthropic/claude-sonnet-4.5",
  "provider": "Anthropic",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": { "role": "assistant", "content": "Hello! How can I help you today?" }
    }
  ],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 9,
    "total_tokens": 21,
    "prompt_tokens_details": { "cached_tokens": 0 },
    "completion_tokens_details": { "reasoning_tokens": 0 },
    "cost": 0.000171
  }
}

usage.cost is the cost of the request in USD. api14 debits it from your balance after the request finishes. The id is the generation id, which you can look up later with Generation lookup.

Billing notes

You pay the model's listed price; the charge is returned in usage.cost. See Pricing & billing.
You are not charged for failed requests (any 4xx or 5xx status).
If the request includes models: [...], every fallback model must be allowed for your key and plan, or the gateway rejects the request before it is sent upstream.

Errors

Status	Code	When
400	`model_required`	`model` is missing.
400	`invalid_json`	The body is not a JSON object.
401	`invalid_api_key`	The key is missing, invalid or revoked.
402	`insufficient_balance`	Your USD balance is $0.
403	`model_not_allowed` / `plan_required`	Your key's allowlist or your plan does not cover this model.
404	`model_not_found`	Unknown or disabled model id.
413	`payload_too_large`	The request body is larger than 25 MB.
429	`rate_limited` / `daily_limit_reached`	RPM limit or the key's daily USD limit was reached.
502	`upstream_unavailable`	The upstream provider could not be reached.

Upstream errors such as context length exceeded or content moderation come back with their original status code and the same error shape. See Errors.