‘AI 속도조절’ 주장한 앤트로픽·오픈AI, ‘가성비 경쟁’ 맞불

Two major AI companies, Anthropic and OpenAI, have entered the race for 'affordability' with their new models, intensifying the competition. Both companies are planning to go public later this year and next, with Anthropic unveiling its new AI model 'Claude Opus 5.5' on the 22nd. Claude Opus 5.5 outperforms Anthropic's top model 'Claude Pillar 5.1' in most tasks, while cutting costs by 40% compared to its predecessor (Opus 5) and speeding up output by over 30%.

The company claims the model excels in coding agents, computer usage, and expert tasks. In fact, it topped the global AI performance ranking 'Intelligence Index' (AAII) with 58 points, surpassing the combined top score of 53 points by Google's 'GPT-6 Astro'. Anthropic also revealed the model is part of their safety assessment, testing the AI's behavior through around 2,000 scenarios during the alignment process to ensure it aligns with human intent.

OpenAI, on the other hand, showcased its lower-level models 'GPT-6 Sol' and 'GPT-6 Luna', focusing on their usability at a significantly lower cost compared to their predecessor, GPT-5.6. OpenAI highlighted that 'GPT-6 Astro' was a groundbreaking AI model, while 'GPT-6 Sol' and 'GPT-6 Luna' bring down the cost of AI benefits to a wider audience.

Recent debates around the 'AI speed-control theory' have intensified as both companies gear up for more aggressive competition. Anthropic is reportedly considering another next-generation model release before its own IPO, while OpenAI plans to unveil multiple new products at its annual developer event 'Devgay 2026' on the 29th. According to Financial Times, US leading AI research institutions are moving away from just aiming for the top model performance and focusing on cost sensitivity to secure corporate clients.