Recently, a Chinese quantitative company called Jiukun Investment launched a new AI large model - the IQuest-Coder-V1 series, which immediately sparked discussions in the tech circle. This model, with its powerful code generation and understanding capabilities, quickly made it to the SWE-Bench Verified list, with the 40B parameter version achieving an excellent score of 81.4%, surpassing many well-known industry models such as Claude Opus-4.5 and GPT-5.2.

The IQuest-Coder-V1 series is designed to meet different development needs, offering three different scale versions: 7B, 14B, and 40B, suitable for code generation, understanding, and software engineering tasks. The 40B version also introduces a Loop variant, aimed at improving the efficiency of parameter utilization. According to the official statement, the model can run on a single consumer-level 3090 or 4090 graphics card, greatly lowering the usage threshold for ordinary developers.

image.png

The model's highlights are not only reflected in its performance but also in its unique training strategy. The IQuest-Coder-V1 adopts a "multi-stage training with code flow," learning from the evolution process of code, allowing the model to better capture the dynamic changes in software logic. This method is completely different from traditional static code training and is more in line with real development scenarios.

The official demonstration of multiple cases has impressed people, from building particle text animations to real-time pixel sandbox games, the IQuest-Coder-V1 has performed effortlessly, greatly enriching the creative possibilities for developers. The success of this model has sparked extensive discussions online, with many people expressing surprise at the Jiukun Investment team, speculating about which top talents they have absorbed.

image.png

Although the model's performance is impressive, the official also reminds users that the generated code still needs to be verified in a sandbox environment, and the model itself does not have the ability to execute code. In addition, as the attention increases, the public has shown great interest in Jiukun Investment's background. The company has been focusing on quantitative investment since its establishment in 2012 and has now become a leading player in domestic quantitative private equity.

image.png

Whether it comes to technological innovation or commercial potential, the release of IQuest-Coder-V1 marks another important progress in China's AI field, and the future application scenarios will be even broader.

Address: https://iquestlab.github.io/#/