← All models

Kimi-Dev-72B

Published Publisher Moonshot AI

Model Positioning

Kimi-Dev-72B is an open-source code large language model for software engineering problem solving. It is based on Qwen2.5-72B and has been trained on code-related data and optimized through reinforcement learning. The official announcement was published on June 16, 2025. (moonshotai.github.io)

Capabilities and Uses

  • Locates files that need modification based on problem descriptions, generates code fixes, and writes unit tests to reproduce issues.
  • Can assist with addressing software defects in code repositories, generating patches and tests, and software engineering research.
  • The official report states that it achieved 60.4% on SWE-bench Verified. This result used test-time self-play, generating up to 40 patch candidates and 40 test candidates per problem, and should not be regarded as a single-generation success rate. (moonshotai.github.io)
  • The maximum position length in the model configuration file is 131,072 tokens; the actual usable length depends on the deployment configuration. (huggingface.co)

Weights and Deployment

The official publisher has provided model weights and project code. The model card specifies the MIT license, and the model can be downloaded and self-deployed. The model repository name moonshotai/Kimi-Dev-72B is a weights identifier, not a verified official API model ID. (huggingface.co)

Limitations

Generated patches and tests may be incorrect, and benchmark performance does not guarantee effectiveness across all repositories or tasks. Code should be reviewed and relevant tests run in an isolated environment before use. The official evaluation used multiple candidate generations; self-deployment and reproduction of this process require corresponding computing resources. (moonshotai.github.io)

Pricing

As of November 18, 2026, searches of the official domestic pricing page in Chinese yuan (CNY) and the official international pricing page in US dollars (USD) did not verify an official input or output token price specifically applicable to Kimi-Dev-72B; billing amounts and units are therefore unverified. Prices for other Kimi models cannot be applied to this model. Self-deployment may still incur computing and infrastructure costs. (platform.kimi.com)

Official Sources