Artificial intelligence programming tools are experiencing a new round of speed improvements. On June 15, Moonshot AI announced the official launch of the high-speed version of its Kimi K2.7Code model. This service is now available to Kimi Code Beta program members, API developers, and Kimi Business users.
The core of this newly launched "high-speed version" lies in a significant leap in response efficiency. According to the official introduction, the model logic remains consistent with the previous Kimi K2.7Code, but through technical optimization, the output speed has increased by 5 to 6 times. In practical programming scenarios, the output speed for short context tasks can reach up to 260 Tokens per second, while regular programming tasks (taking the median input length) can maintain a stable output of around 180 Tokens per second.
To achieve this efficiency upgrade, users need to make corresponding trade-offs in price. The high-speed version is priced twice as much as the standard version of Kimi K2.7Code. The specific billing standards are: the standard input and output prices are 13 yuan and 54 yuan per million Tokens, respectively. If the cache is hit, the input price is 2.6 yuan per million Tokens.
Kimi K2.7Code was officially released on June 12, positioning itself as a dedicated model for long-context programming tasks. This series of models has shown significant improvements in instruction following ability and long-range programming tasks, especially optimizing the issue of excessive thinking when handling complex code logic, reducing the average token consumption by about 30%.
For developers and enterprise users, the addition of this high-speed version means that they can achieve shorter response times by investing more costs while ensuring high-quality model output. This is undoubtedly an important efficiency tool option for programming workflows that require frequent code iteration and real-time interaction.