TRAE.ai announces that its custom model service provider has added Kimi, officially launching the Kimi-K2 model and supporting the supermodel Grok-4 (Beta) in the international version. This innovative move aims to meet the growing demand for diversity and professionalism among developers.

The powerful performance of the Kimi-K2 model

Kimi-K2 is a foundational model based on a mixture of experts (MoE) architecture, featuring excellent code capabilities and general agent task processing abilities. In tasks such as code generation and mathematical reasoning, Kimi-K2 demonstrates leading performance. Especially in front-end development, it can generate high-quality code that combines design sense and visual expression.

According to feedback from some third-party users, the overall performance of Kimi-K2 has surpassed Claude3.7Sonnet and DeepSeek V3, and is approaching the level of GPT-4.1, Claude4, and Gemini2.5pro.

image.png

Simple three steps to access Kimi-K2

To integrate Kimi-K2 into your workflow, users just need to follow these simple steps:

image.png

  1. Open the Kimi Open Platform and create an API Key (link: https://platform.moonshot.cn/console/api-keys).
  2. Open the model list in TRAE and select Add Model.
  3. Select the service provider Kimi, the model Kimi-K2, and enter the API Key to complete the integration.

New addition of Grok-4 (Beta) in the international version

In addition to Kimi-K2, the TRAE international version also launched the supermodel Grok-4 (Beta), which is still in the testing phase. Pro users may experience brief waiting times when using it.

image.png

A comprehensive range of models

The model lineup in the TRAE international version includes mainstream large models globally, providing full coverage from complex tasks to efficient scenarios. The supermodel section currently includes:

  • Claude-4-Sonnet (Beta)
  • Claude-3.7-Sonnet
  • Claude-3.5-Sonnet
  • Grok-4 (Beta)👈 New addition
  • Gemini-2.5-Pro
  • GPT-4.1
  • GPT-4o
  • DeepSeek-V3-0324

  • DeepSeek-Reasoner (R1)

    ➤ TRAE China Version: https://www.trae.cn

    ➤ TRAE International Version: https://www.trae.ai