Count tokens

POST /v1/chat/count_tokens returns the number of input tokens a chat request would use, without running inference. Use it for cost estimates, context-window checks, and validating a request before you queue it.

The count matches the prompt_tokens you would be billed for on the same request to /v1/chat/completions. It accounts for the system prompt, every message, tool definitions, the model's chat template, and the thinking-mode tokens added by reasoning_effort.

curl -s "https://api.privatemind.com/v1/chat/count_tokens" \
  -H "Authorization: Bearer $PMIND_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "fast",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user",   "content": "Hello, how are you today?"}
    ]
  }'
# => { "input_tokens": 23, "context_window": 131072 }

import httpx

resp = httpx.post(
    "https://api.privatemind.com/v1/chat/count_tokens",
    headers={"Authorization": f"Bearer {PMIND_KEY}"},
    json={
        "model": "fast",
        "messages": [
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user",   "content": "Hello, how are you today?"},
        ],
    },
)
print(resp.json()["input_tokens"])

const resp = await fetch('https://api.privatemind.com/v1/chat/count_tokens', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${PMIND_KEY}`,
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'fast',
    messages: [
      { role: 'system', content: 'You are a helpful assistant.' },
      { role: 'user', content: 'Hello, how are you today?' },
    ],
  }),
});
const { input_tokens } = await resp.json();
console.log(input_tokens);

Parameters

The request takes the token-affecting fields of a chat completion. Sampling parameters (temperature, max_tokens, stream, …) are accepted but ignored: they do not change the input token count.

• model (required): id or alias of a chat model, resolved exactly as /v1/chat/completions resolves it. Unknown models return model_not_found.
• messages (required): the conversation, same shape as chat completions.
• tools: tool definitions. Counted for models whose chat template renders tools into the prompt.
• reasoning_effort: off | low | medium | high. On a thinking model this changes the template and therefore the count, so pass the same value you will send to the completion.

Response

{
  "input_tokens": 23,
  "context_window": 131072
}

• input_tokens: input tokens the request would consume, the same number billed as prompt_tokens on the matching completion.
• context_window: the model's maximum context length in tokens. Compare input_tokens (plus the completion you expect to generate) against this to detect overflow before you send.

Counting is free: a count_tokens call is never billed and writes no usage.

Cost and context checks

Two common uses:

• Cost estimate. Multiply input_tokens by the model's input rate (see Usage for per-model rates) to show an estimated cost before sending.
• Context-window guard. If input_tokens already approaches context_window, trim or summarise the conversation before calling /v1/chat/completions, rather than letting the model reject an over-long prompt.

Where next

• Chat completions: the endpoint whose prompt_tokens this mirrors.
• Models: find a model's id, alias, and context window.
• Usage: per-model token rates for turning a count into a cost.
• Rate limits & budgets: how budgets and RPM apply to your keys.