MiniMax M3 Launches with Game-Changing Features and Industry-Leading Performance
On June 1, 2026, MiniMax unveiled its latest AI model, MiniMax-M3, a large language model that is already making waves across the global AI ecosystem. The M3 model breaks new ground with a 1-million-token context window and native support for multimodal inputs — including text, images, audio, and video, alongside integrated computer vision capabilities. This launch marks a significant milestone not only for MiniMax but also for the competitive landscape of advanced AI models, as the M3 demonstrably outperforms industry stalwarts GPT-5.5 and Gemini 3.1 Pro on key agentic benchmarks, including SWE-Bench Pro, Terminal Bench 2.1, and BrowseComp. MiniMax has also announced that open weights will be released within ten days, making the model freely downloadable and customizable for enterprises.
Industry observers and analysts alike are taking note, with VentureBeat highlighting the model’s impressive benchmark results combined with its disruptive pricing strategy. At launch, MiniMax-M3 is available via API at a special introductory price of $0.30 per million input tokens and $1.20 per million output tokens, with full pricing of $0.60/$2.40 per million tokens, which still represents just 8–20% of the cost of leading proprietary US models. A subscription plan starts at $20 per month.
A 1 Million Token Context Window and Native Multimodality
The 1 million token context window is perhaps the most immediately striking feature of the MiniMax M3, enabling unprecedented length and complexity in AI interactions. This vast context capacity allows the model to maintain coherent understanding and generate responses across extensive conversations, documents, or multimedia inputs without losing context or requiring costly chunking strategies. This is a critical advancement for applications that demand long-form reasoning, intricate problem-solving, or continuous multi-turn dialogue.
Complementing this is M3’s native multimodal input capability, supporting seamless integration of text, images, audio, and video within a single query. This capability opens up significant new use cases across industries—from interactive educational content and sophisticated multimedia search to advanced medical diagnostics and real-time multimedia translation. By natively understanding and processing multiple data formats simultaneously, the M3 model sets a new standard for AI versatility and intelligence.
Further extending its utility, the M3 incorporates computer capabilities, allowing it to perform actions that involve interaction with external software or hardware. This positions the model not simply as a passive knowledge engine but as an active agent capable of executing tasks or interfacing with complex systems dynamically, a feature increasingly vital for enterprise adoption and automation.
Outperforming GPT-5.5 and Gemini 3.1 Pro on Agentic Tasks
The M3’s performance on key agentic benchmarks confirms its technical competitiveness. On SWE-Bench Pro, an autonomous software engineering benchmark, M3 scores 59.0%, ahead of both GPT-5.5 and Gemini 3.1 Pro. On Terminal Bench 2.1, which evaluates command-line execution accuracy, M3 achieves 66.0%. On MCP Atlas, it scores 74.2%, and on BrowseComp, which tests autonomous browsing and information retrieval, M3 scores 83.5%, outstripping Claude Opus 4.7’s score of 79.3.
It is worth noting that Anthropic’s newly released Claude Opus 4.8 leads M3 on several of these same benchmarks: 69.2% on SWE-Bench Pro, 74.6% on Terminal Bench 2.1, and 83.4% on OSWorld-Verified. The competitive picture is therefore nuanced: M3 outperforms the previous generation of frontier models at a fraction of the cost, but the very latest closed-source premium models from Anthropic maintain an edge on the most demanding tasks.
Cost Disruption: High Performance at a Fraction of the Price
One of MiniMax’s most significant strategic advantages lies in the M3’s pricing structure. At a full price of $0.60 per million input tokens and $2.40 per million output tokens — and currently available at an introductory $0.30/$1.20, M3 is priced at roughly 8–20% of comparable offerings from OpenAI and Google. Even at full price, GPT-5.5 costs $5.00/$30.00 per million tokens and Claude Opus 4.8 costs $5.00/$25.00, making M3 dramatically cheaper for comparable agentic performance. This aggressive pricing not only lowers the barrier for startups, developers, and enterprises to access cutting-edge AI but also threatens to reshape the competitive dynamics of the AI cloud services market.
By combining strong agentic performance with cost efficiency, MiniMax is positioning itself as a disruptive force capable of capturing substantial market share. The M3’s affordability enables broader adoption across sectors where AI costs have been a limiting factor, including education, healthcare, media, and industrial automation. The forthcoming open-weights release will further democratize access, allowing enterprises to download, fine-tune, and self-host the model at zero API cost.
Market Reaction: MiniMax Shares Soar Amid Dual Listing Plans
MiniMax’s commercial trajectory has mirrored its technological achievements. Since its January 2026 IPO on the Hong Kong Stock Exchange, MiniMax shares have surged over 400%, reflecting investor confidence in the company’s prospects and its AI breakthroughs. This remarkable stock performance underscores the market’s recognition of MiniMax’s strategic positioning and growth potential.
Looking ahead, MiniMax is preparing a dual listing on China’s STAR Market, the Shanghai Stock Exchange’s high-tech board focused on innovation-driven enterprises. This move aims to deepen MiniMax’s capital market presence and enhance liquidity while fostering closer alignment with China’s burgeoning AI ecosystem. The EastFrontier report provides detailed insights into the company’s dual listing plans and the broader implications for its market expansion.
MiniMax’s M3 launch arrives amid intensifying global competition in AI development, particularly between China and the United States. The model’s combination of advanced multimodality, extended context capacity, and computer use capability marks a leap forward for Chinese AI innovation. It reflects the country’s growing ability to produce world-class AI technologies that rival or surpass Western counterparts, not just in raw performance but also in cost-effectiveness and practical application.
This development has several strategic dimensions:
•Technological Sovereignty: By advancing indigenous AI models like M3, China strengthens its autonomy in critical digital infrastructure, reducing reliance on foreign AI providers and mitigating geopolitical risks.
•Enterprise and Consumer Adoption: The affordability and versatility of M3 lower entry barriers for Chinese enterprises and startups, accelerating AI integration across industries and potentially driving new waves of productivity and innovation.
•Global Market Disruption: MiniMax’s pricing and performance could disrupt global AI service markets, forcing Western incumbents to reevaluate pricing strategies and innovation roadmaps.
•Talent and Collaboration: The prominence of MiniMax in international benchmarks and markets enhances China’s appeal as a global AI hub, attracting talent, investment, and partnerships.
The release of the MiniMax M3 model signals a shift toward more capable, cost-effective, and versatile AI systems that are better suited to real-world, multimodal challenges. As enterprises and developers begin to integrate M3’s capabilities, we can expect accelerated innovation cycles and new AI-driven applications across sectors.
Moreover, MiniMax’s strategic moves, capitalizing on both Hong Kong and Shanghai’s STAR Market listings, position it to fuel further R&D and product development, potentially ushering in a new era of AI leadership from China. For observers of the AI landscape, the M3 launch is a pivotal event that underscores the shifting balance of power and the expanding possibilities of artificial intelligence worldwide.
For more on MiniMax’s market performance and strategic developments, see our detailed coverage on EastFrontier.