1美元时薪？这才是打工人的「梦中情模」

Core Viewpoint - The release of MiniMax M2.5 presents a cost-effective alternative to Claude Opus 4.6, which has seen a significant price increase, creating a dilemma for users between high intelligence and affordability [1][47]. Pricing and Cost Comparison - Claude Opus 4.6's "Turbo" version has increased its cost from $25 to $150 per million tokens, a 6-fold increase [1]. - In contrast, MiniMax M2.5 offers a much lower cost structure, allowing users to operate at $1 per hour for 100 tokens per second, and $0.3 per hour for 50 tokens per second [13][14]. Performance Metrics - MiniMax M2.5 has achieved industry-leading performance in various benchmarks, including Multi-SWE-Bench and SWE-Bench Verified, outperforming Opus 4.6 in specific tasks [2][4]. - In internal evaluations, M2.5 autonomously completed 30% of tasks and generated 80% of new code submissions, showcasing its efficiency [6]. Technical Specifications - MiniMax M2.5 is notable for its compact size, with only 10 billion parameters, making it the smallest flagship model in its class [8]. - The model's throughput can reach 100 TPS, and its end-to-end task execution time has improved from an average of 31.3 minutes to 22.8 minutes, matching Opus 4.6's speed [13]. Development and Training Innovations - MiniMax has rapidly evolved, releasing three versions (M2, M2.1, M2.5) in just 108 days, demonstrating the fastest progress in the industry compared to competitors [39][41]. - The training framework, Forge, allows for efficient training and task execution by decoupling the model from specific agent products, enhancing its adaptability [43]. Market Implications - The affordability and performance of MiniMax M2.5 challenge the notion that high intelligence must come with high costs, potentially democratizing access to advanced AI tools [47][48]. - The combination of MiniMax M2.5 and its desktop agent could provide a powerful tool for professionals, enhancing productivity without significant financial burden [48].