OpenAI and Anthropic Launch New AI Models to Cut Operational Costs

OpenAI and Anthropic have both introduced their next-generation AI models in simultaneous launches, marking a move of the industry towards efficiency and speed. 

OpenAI rolled out the GPT-6 Sol and GPT-6 Luna mid-range models as part of the GPT-6 suite, whereas Anthropic released Claude Opus 5.5 to handle sophisticated software engineering and long-term operations.   

New architecture of OpenAI is said to cut API inference costs in half from the last iteration, giving GPT-6 Sol at a cost of $2 per million input tokens and $10 per million output tokens. 

The less costly GPT-6 Luna model has reduced launch costs to $0.10 per million input tokens. Meanwhile, Anthropic says that Claude Opus 5.5 reduces the expenses of workloads by 40% from Opus 5 and boosts output generation speeds up to 2.5x faster in fast modes.

Both these AI developers created their models based on agentic computer usage, large-file parsing, and coding processes that can occur autonomously. Through maximizing the caching of memory and minimizing the number of tokens needed for long projects with multiple steps, both aim to sustain the enterprise software workloads without breaching budgets. 

This competition is indicative of a change in the entire industry that sees moving away from benchmark performance superiority to unit economics. The concurrent launches by OpenAI and Anthropic models represent a defining moment in artificial intelligence research in that technological advancement is now matched with affordability. 

The reduction in the cost of inference allows the researchers to develop sophisticated, multi-agent systems without incurring high overhead costs.