GPT-6.1 Terra vs. Claude Opus 5.5: Cheaper and faster in three tests at max (thenewstack.io)

OpenAI has recently launched GPT-6.1 Terra, which has shown promising performance comparisons against Anthropic's Claude Opus 5.5 in three critical tests focused on coding and data analysis tasks. Terra scored similarly to Opus, achieving 60% on the Agents’ Last Exam while costing just $1.01 per task compared to Opus's $8.82. Notably, Terra demonstrated over three times the token efficiency of Opus while maintaining competitive accuracy, making it a significant advancement in cost-effective AI capabilities. The implications of this release are substantial for the AI/ML community, particularly in academic and industrial applications where budget constraints are crucial. Terra not only performed on par but often excelled in terms of speed and token usage efficiency. For instance, in a cell lineage tracking test, Terra outperformed Opus on accuracy while costing substantially less—$3.47 versus Opus's $82.91. These results suggest that Terra could be a more viable option for researchers and developers looking for robust AI solutions without the financial burden typically associated with high-performance models. As AI continues to permeate various sectors, advancements like Terra’s cost-efficiency and speed could democratize access to powerful data processing tools.