GPT-6.1 Sol Replaces GPT-6 Sol After Seven Days With Near-Astra Intelligence

GPT-6.1 Sol Replaces GPT-6 Sol After Seven Days With Near-Astra Intelligence

GPT-6.1 Sol replaces GPT-6 Sol only seven days after its release. It scores one point below GPT-6 Astra on the Intelligence Index while costing less than one quarter as much per task.

GPT-6.1 Sol uses the same headline pricing as GPT-6 Sol: $2 per million input tokens and $10 per million output tokens. However, the cache-read discount increases from 90% to 95%. As a result, its blended price for agentic workloads is slightly lower than GPT-6 Sol's. This follows GPT-6 Sol's original 50% price reduction compared with GPT-5.6 Sol.

Key findings

• Near-Astra Intelligence: GPT-6.1 Sol gains four points on the Intelligence Index compared with GPT-6 Sol and five points compared with GPT-5.6 Sol, placing it one point below GPT-6 Astra. It improves by four points on AA-Briefcase v1.1 and five points on GDPval-AA v2.1. Other gains include a 12-point increase on Terminal-Bench 4.0, a five-point increase on Humanity's Last Exam, a six-point increase on GDP.pdf, and an eight-point increase in AA-Omniscience Accuracy. Its hallucination rate falls from 60% to 54%.
• Cost efficiency: At maximum effort, GPT-6.1 Sol costs $0.72 per Intelligence Index task, compared with $3.26 for GPT-6 Astra. That is 31% less than GPT-6 Sol at $1.05 and 64% less than GPT-5.6 Sol at $1.99. At every effort level, GPT-6.1 Sol extends the cost-efficiency Pareto frontier: for a given intelligence score, no less expensive model is listed.
• Token efficiency: GPT-6.1 Sol produces approximately 10% to 30% more output tokens than GPT-6 Sol across effort levels. Its higher Intelligence Index scores nevertheless make the low- and medium-effort settings Pareto optimal for token efficiency.
• Coding Agent Index: At maximum effort, GPT-6.1 Sol gains three points over GPT-6 Sol and remains two points below GPT-6 Astra.

Coding Agent Index

GPT-6.1 Sol occupies the lower-price portion of the Pareto frontier for the Artificial Analysis Coding Agent Index compared with cost per task. At the xhigh effort setting, it scores one point higher than GPT-6 Astra while costing less than 15% as much per task. This is a six-point improvement over GPT-6 Sol at maximum effort.

The xhigh setting outperforms the maximum-effort setting by three points.

Token efficiency

Across the Intelligence Index effort settings, GPT-6.1 Sol uses 10% to 30% more output tokens than GPT-6 Sol. Its intelligence gains mean that the low- and medium-effort settings remain Pareto optimal for token efficiency.

AA-Omniscience

At maximum effort, GPT-6.1 Sol increases its AA-Omniscience Accuracy score by eight points. At the same time, its hallucination rate decreases by six points, from 60% to 54%.

AA-Briefcase

GPT-6.1 Sol improves by approximately 80 Elo in AA-Briefcase. The increase is driven by higher rubric and Analytical Quality Elo scores, while Presentation Elo declines slightly.

Results by evaluation

The Artificial Analysis Intelligence Index v4.3.2 provides a breakdown of GPT-6.1 Sol's results across its individual evaluations. The results show broad improvements over GPT-6 Sol, including gains in agentic knowledge work, coding, factual reliability, and several of the component benchmarks used to calculate the overall index.