# Chat completions endpoint: POST /api/v1/chat/completions > Chat completions endpoint reference: every body field of POST /api/v1/chat/completions, the response, finish reasons and token usage. OpenAI-compatible. - URL: https://llmwithcrypto.com/docs/chat-completions/ - Last updated: 2026-10-23 - Publisher: LLMwithCrypto (https://llmwithcrypto.com) One endpoint answers every model. It takes the same body as the OpenAI chat completions API and returns the same object, so an existing integration needs a new base URL, a new key and a model id from our list. **Quick answer — Which fields does the chat completions endpoint of LLMwithCrypto accept?** It accepts the OpenAI chat completions body: `model` and `messages` are required; `max_tokens`, `temperature`, `top_p`, `stop`, `stream`, `stream_options`, `tools`, `tool_choice` and `response_format` are optional. **LLMwithCrypto** returns a `chat.completion` object with a `usage` count, billed per token from your credit, as documented on 23 October 2026. Tool calling and JSON output depend on the model. ## The endpoint at a glance | Attribute | Value | | --- | --- | | Method and URL | `POST https://llmwithcrypto.com/api/v1/chat/completions` | | Authentication | Pass your API key as a Bearer token in the Authorization header, as the OpenAI SDK does | | Body | JSON, `Content-Type: application/json` | | Response | `chat.completion` object, or server-sent events with `stream: true` | | Billing | You pay per token from your balance: input and output tokens are counted by the model’s own tokenizer and charged at that model’s price per 1M tokens. A call that ends in an error costs nothing | | Features | Streaming works on every model; tool calling and JSON output follow what each model itself supports | ## Create a chat completion `POST /api/v1/chat/completions` Sends a conversation to one model and returns its answer with the token counts. The example asks DeepSeek V4 Pro a short question with a 200-token cap; ids, times and counts in the response are illustrative. Authentication: API key in the `Authorization: Bearer API_KEY` header. **Request body** | Field | Type | Required | Description | | --- | --- | --- | --- | | `model` | string | Yes | The model id, such as `deepseek-v4-pro` or `gpt-oss-120b`. `GET /api/v1/models` returns every id your key can call. | | `messages` | object[] | Yes | The conversation, oldest first. Each item has a `role` (`system`, `user`, `assistant` or `tool`) and a `content` string; `tool` messages also carry a `tool_call_id`. | | `max_tokens` | integer | No | The most tokens the answer may use. Without it, the model may write until its context window is full, and you pay for every token it writes. | | `temperature` | number | No | 0 to 2. Lower is more predictable, higher more varied. The default depends on the model. | | `top_p` | number | No | 0 to 1, nucleus sampling. Change `temperature` or `top_p`, not both. | | `stop` | string or string[] | No | Up to 4 sequences that end the answer when the model writes them. They are not included in the answer. | | `stream` | boolean | No | `true` sends the answer as server-sent events, piece by piece. See [streaming](https://llmwithcrypto.com/docs/streaming/). | | `stream_options` | object | No | `{"include_usage": true}` adds a last event with the `usage` of a streamed call. | | `tools` | object[] | No | Functions the model may ask you to call, in the OpenAI format. Works on the models that support tool calling. See [tool calling](https://llmwithcrypto.com/docs/tool-calling/). | | `tool_choice` | string or object | No | `auto` (default when `tools` is set), `none`, `required`, or `{"type": "function", "function": {"name": "…"}}`. | | `response_format` | object | No | `{"type": "json_object"}` asks for a valid JSON object, on the models that support it. Say in your prompt what the JSON should contain. | *Example: a short answer from DeepSeek V4 Pro* ```json { "model": "deepseek-v4-pro", "messages": [ { "role": "system", "content": "Answer in two sentences." }, { "role": "user", "content": "What is a context window?" } ], "max_tokens": 200 } ``` *Example: a JSON object from gpt-oss-120b, low temperature* ```json { "model": "gpt-oss-120b", "messages": [ { "role": "user", "content": "Return {\"city\": string, \"country\": string} for the Eiffel Tower." } ], "response_format": { "type": "json_object" }, "temperature": 0.2, "max_tokens": 100 } ``` **Response** `200` ```json { "id": "chatcmpl-8f2c41d0a7b94e1c", "object": "chat.completion", "created": 1792742400, "model": "deepseek-v4-pro", "choices": [ { "index": 0, "message": { "role": "assistant", "content": "A context window is the maximum number of tokens a model can read and write in one request. Anything beyond it must be cut or summarized before the call." }, "finish_reason": "stop" } ], "usage": { "prompt_tokens": 27, "completion_tokens": 34, "total_tokens": 61 } } ``` *curl* ```bash curl -s "https://llmwithcrypto.com/api/v1/chat/completions" \ -H "Authorization: Bearer $LLMWITHCRYPTO_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "deepseek-v4-pro", "messages": [ {"role": "system", "content": "Answer in two sentences."}, {"role": "user", "content": "What is a context window?"} ], "max_tokens": 200 }' ``` *Python* ```python import os from openai import OpenAI client = OpenAI( base_url="https://llmwithcrypto.com/api/v1", api_key=os.environ["LLMWITHCRYPTO_API_KEY"], ) resp = client.chat.completions.create( model="deepseek-v4-pro", messages=[ {"role": "system", "content": "Answer in two sentences."}, {"role": "user", "content": "What is a context window?"}, ], max_tokens=200, ) print(resp.choices[0].message.content) print(resp.usage.prompt_tokens, "in,", resp.usage.completion_tokens, "out") ``` *Node.js* ```javascript import OpenAI from 'openai'; const client = new OpenAI({ baseURL: 'https://llmwithcrypto.com/api/v1', apiKey: process.env.LLMWITHCRYPTO_API_KEY, }); const resp = await client.chat.completions.create({ model: 'deepseek-v4-pro', messages: [ { role: 'system', content: 'Answer in two sentences.' }, { role: 'user', content: 'What is a context window?' }, ], max_tokens: 200, }); console.log(resp.choices[0].message.content); console.log(resp.usage.prompt_tokens, 'in,', resp.usage.completion_tokens, 'out'); ``` ## How do I write the messages? Send the whole conversation every time, oldest message first: the API keeps no memory between calls, and nothing of your prompts is stored. We do not store your prompts or the model’s answers: once a response is sent, its text is gone. Each call leaves one log line with the time, the account, the model and the token counts, deleted after 30 days. - **`system`**: instructions that frame every answer, such as tone, language or output format. Put one at the top. - **`user`**: what the person or your program asks. - **`assistant`**: the model’s earlier answers, so it can continue the conversation. - **`tool`**: the result of a function the model asked you to run, with the `tool_call_id` it gave you. Every message counts as input tokens, so long histories cost more on each turn. Trim or summarize old turns when a conversation grows. ## How long can a request be? Input and output together must fit in the model’s context window. As of 23 October 2026, DeepSeek V4 Pro reads up to 1M tokens, gpt-oss-120b up to 128K and Qwen3 235B up to 256K; the [models page](https://llmwithcrypto.com/models/) lists all of them. A request that does not fit comes back as `400` with the code `context_length_exceeded` and costs nothing. Set `max_tokens` so that the prompt plus the longest answer you accept stays inside the window. ## What does finish_reason tell me? | `finish_reason` | Meaning | What to do | | --- | --- | --- | | `stop` | The model ended its answer, or wrote one of your `stop` sequences | Nothing: the answer is complete | | `length` | The answer reached `max_tokens` or the context window | Raise `max_tokens`, or ask for a shorter answer | | `tool_calls` | The model wants you to run one or more functions | Run them and send the results back, see [tool calling](https://llmwithcrypto.com/docs/tool-calling/) | ## Which OpenAI parameters are not supported? The fields in the table above are the whole contract. Others from the OpenAI API, such as `n` (several answers per call), `logprobs`, `audio` or image inputs, are not supported: send only the fields listed here. The models are text models; none of them takes images through this endpoint. The `model` field only takes the ids on our list. We do not sell access to GPT, Claude, Gemini or Grok: their makers forbid reselling API access without their approval, and we do not have it. ## Chat completions: common questions ### Is there a legacy completions endpoint? No. Only the chat format exists; a single prompt is a `messages` array with one `user` message. ### What does a call cost? You pay per token from your balance: input and output tokens are counted by the model’s own tokenizer and charged at that model’s price per 1M tokens. A call that ends in an error costs nothing. Per-model prices are on the [pricing page](https://llmwithcrypto.com/pricing/). ### Can I see my past requests? No, by design. We do not store your prompts or the model’s answers: once a response is sent, its text is gone. Each call leaves one log line with the time, the account, the model and the token counts, deleted after 30 days. ### Why did my answer stop in the middle? `finish_reason` is then `length`: the answer hit `max_tokens`. Raise it, within the model’s context window. --- LLMwithCrypto: LLMwithCrypto is an OpenAI-compatible LLM API for open-weight models — DeepSeek V4 Pro, DeepSeek V4.1 Flash, DeepSeek R1, Qwen3.5 397B A17B, Qwen3 235B A22B Instruct, gpt-oss-120b, gpt-oss-20b, GLM-4.7, Llama 3.3 70B Instruct, Gemma 4 31B, Mistral Large 3, Mistral Small 4 — paid per token from a prepaid crypto balance, at 5% below the median price of the public hosts. No KYC: no ID, no selfie, no documents. Prompts and outputs are not stored.