Chinese Startup Claims AI Coding Capabilities Exceed Those of GPT-5.5: A Detailed Examination

MiniMax Launches New AI Model, M3, Claiming Superior Performance

Chinese artificial intelligence startup MiniMax has launched its new flagship AI model, M3, which the company claims outperforms OpenAI's GPT-5.5 and Google's Gemini 3.1 Pro on the SWE-Bench Pro coding benchmark.

According to MiniMax, M3 can process up to one million tokens in a single context window, enabling it to handle large software projects, lengthy documents, and complex coding workflows. The launch marks MiniMax's most ambitious move yet as it seeks to challenge leading global AI developers.

Based in Shanghai, the company described M3 as a coding-focused model designed for software engineering, AI agents, and automated workflows. It also supports multimodal inputs, allowing users to work with text, images, and video in a single system.

Read also: Google Pixel Watch Leak: Diver's Discovery Sparks Speculation

M3's Architecture and Performance

MiniMax reported that the new model uses a redesigned architecture known as MiniMax Sparse Attention, which processes only the most relevant information within large datasets. This approach significantly reduces computing requirements while improving speed when working with long contexts.

M3 surpassed GPT-5.5 and Gemini 3.1 Pro on SWE-Bench Pro, a benchmark created to test AI systems on complex software engineering tasks that resemble real-world programming challenges. The benchmark is widely regarded as a demanding measure of coding performance because it evaluates an AI model's ability to solve large and realistic software problems.

Read also: Google Unveils Gemini Omni: AI-Powered Video Creation from Text Input

Cost and Accessibility

MiniMax said M3 is priced well below many leading proprietary models, a strategy that could attract developers and businesses looking for advanced AI capabilities at lower operating costs. The company is also positioning M3 as an open weight model, allowing broader access for researchers and developers.

Industry Reaction

The launch of M3 comes at a time when competition in artificial intelligence is intensifying worldwide. Major technology companies and startups are racing to develop more capable AI systems that can write software, solve complex problems, and carry out tasks with minimal human input.

With businesses increasingly adopting AI tools to speed up software development and automate workflows, MiniMax is positioning M3 as a challenger to leading models from OpenAI, Google, and other global rivals. The model's performance claims and large context window are expected to attract attention from developers, enterprises, and investors closely tracking the fast-moving AI sector.