OpenAI Introduces GPT-6.1 Terra
OpenAI has officially introduced GPT-6.1 Terra as the mid-tier model in their GPT-6.1 series, available to users across ChatGPT tiers and developers via the API. The model is designed for everyday agentic, coding, and writing work with enhanced capabilities in processing screenshots and on-screen content. This release expands OpenAI's mid-tier model lineup, providing an efficient option for high-volume workloads while sparking discussions about model tiering and pricing strategies. It directly impacts how developers and enterprise users balance performance and cost when selecting AI models for everyday tasks. GPT-6.1 Terra is priced below GPT-6.1 Sol and offers strong multi-modal capabilities for on-screen content comprehension. Community observations note complex behaviors in reasoning token utilization, where 'none' and 'low' reasoning settings can sometimes exhibit unexpected token costs and counts.
## BACKGROUND
Reasoning tokens are intermediate tokens generated by large language models to process complex problems before outputting a final answer. Model tiers like Terra typically occupy a middle ground in performance, balancing the speed of smaller models with the deep capabilities of larger frontier models.