OpenAI launches GPT-6 Sol and Luna as Anthropic releases Claude Opus 5.5

OpenAI and Anthropic announced cutting-edge AI models in rapid succession, intensifying rivalry not only over AI capabilities but also over the expense of deploying increasingly potent systems. OpenAI released GPT-6 Sol and GPT-6 Luna, more affordable options in its GPT-6 lineup, while Anthropic unveiled Claude Opus 5.5, a top-tier model with a focus on coding, autonomous functions, and professional tasks.

The competition now centers on the amount of practical work models can accomplish per dollar spent, rather than just benchmark scores.

OpenAI's GPT-6 Sol and Luna enhance technology from GPT-6 Astra, aiming for swift and economical everyday use. The most significant change is pricing—GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, while GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens. This represents a 50% discount compared to promotional API fees for the corresponding GPT-5.6 models.

GPT-6 Sol and Luna support text and image inputs, have context windows of roughly 1.05 million tokens, and reportedly make about half as many factual errors as GPT-5.6 Sol in real-world conversations where users reported mistakes. They are accessible via OpenAI's API and will be introduced in ChatGPT Work and Codex for selected paid users. GPT-6 Luna is also available for free and trial in the desktop app.

Anthropic introduced Claude Opus 5.5, the inaugural member of its new Claude 5.5 series, which mirrors the performance of the larger Claude Fable 5.1 model on most tasks but at 40% lower operational costs. Claude Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens, with cache reads at $0.20 per million tokens.

The company highlights its superiority in coding, computer use, and professional work evaluations, boasting output generation that is over 30% faster than Claude Opus 5. Safety is a key concern for Anthropic, which reports that Opus 5.5 outperformed its predecessors in automated behavioral and alignment tests and is more resilient to prompt-injection attacks.

It underwent scrutiny from independent AI safety researchers prior to its release and will be deployed with additional safeguards for sensitive fields like cybersecurity and biology. Anthropic plans to roll out Claude Sonnet 5.5 and Claude Haiku 5.5 in the coming weeks. This near-simultaneous model launch underscores the evolving nature of the AI race: while models are becoming more powerful, the competition is increasingly about affordability and practicality for larger-scale deployment.