Kimi K2.7-Code

Connection details

Availability: Published. Check available models before calling.

Create a key in API Keys, check your account credits, and store the key in BULKTOKEN_API_KEY. Install openai for the Python or Node.js example. See Quick Start for setup commands.

Call Kimi K2.7-Code

The examples stream answer text and use the exact BulkToken model ID. For a single JSON response, set stream to false and read choices[0].message.content.

curl -N https://bulktoken.ai/v1/chat/completions \
  -H "Authorization: Bearer $BULKTOKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kimi-k2.7-code",
  "messages": [
    {
      "role": "user",
      "content": "Explain how a hash table works in three sentences."
    }
  ],
  "stream": true
}'

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["BULKTOKEN_API_KEY"],
    base_url="https://bulktoken.ai/v1",
    timeout=120.0,
)

stream = client.chat.completions.create(
    model="kimi-k2.7-code",
    messages=[{"role": "user", "content": "Explain how a hash table works in three sentences."}],
    stream=True,
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.BULKTOKEN_API_KEY,
  baseURL: "https://bulktoken.ai/v1",
  timeout: 120_000,
});

const stream = await client.chat.completions.create({
  model: "kimi-k2.7-code",
  messages: [{ role: "user", content: "Explain how a hash table works in three sentences." }],
  stream: true,
});
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

Use it in your application

Start with text code and test output. For multi-step tool use, keep complete assistant messages, including reasoning and tool calls. The official model accepts tool_choice auto or none; do not assume forced tool selection works.

Example task: Explain this failing test and propose a minimal patch.

Thinking cannot be disabled. Leave sampling parameters at their defaults: temperature 1, top_p 0.95, n 1, and zero presence/frequency penalties. The default output budget is not a maximum.

Inputs and request budgets

These are published model specifications, not a guarantee that every route supports the maximum. Input, output, and any reasoning budget must fit the selected route’s limits. A maximum is not a recommended request size. See the model catalog for current listings; models awaiting channel setup are not callable yet.

The examples here cover text input. Image/video encoding, file-size limits, and reasoning controls are route-specific; verify those options before adding media to a production integration.

Streaming, tools, and JSON

The published model catalog lists streaming, tool calling, and structured output. Start with the minimal call above, then add the feature you need:

• Streaming: consume answer deltas and handle empty chunks and interruption.
• Tool calling: execute tools in your application and return results with matching IDs.
• JSON output: validate responses and check the exact format supported by the route.

Support for an OpenAI-compatible endpoint does not imply support for all provider-specific parameters or every OpenAI API.

Pricing and troubleshooting

BulkToken charges 80% of the official reference rates (20% off). See the model pricing for applicable token categories.

See the model catalog for current input, output, and applicable cache rates. Models awaiting channel setup will appear after publication. Verify actual requests in Usage Logs. Cache discounts apply to reported cache usage, not simply to repeated text.

If a request fails, check the complete key, its model restrictions, your account credits, and the exact model ID kimi-k2.7-code. For 429 or temporary service failures, follow Errors and retries.