Key Takeaways
- • OpenAI launched GPT-5.6 Sol as its flagship frontier model, introducing a tiered family structure (Sol, Terra, Luna) designed to optimize intelligence, cost, and speed across different use cases.
- • GPT-5.6 Sol achieved state-of-the-art performance on presentation quality benchmarks and topped the Agents' Last Exam evaluation with a score of 53.6, outperforming Anthropic's Claude Fable 5 by 13.1 points on long-running professional workflows across 55 fields.
- • The model demonstrates superior cost efficiency: GPT-5.6 Sol achieves performance comparable to Claude Fable 5 at roughly half the estimated cost and 61% faster task completion, while smaller variants (Terra, Luna) deliver Fable 5-level performance at approximately one-sixteenth the cost.
- • OpenAI introduced "ultra," a new high-capability setting that coordinates multiple agents across parallel workstreams to accelerate complex tasks, and enhanced computer use and design judgment capabilities for producing polished, ready-to-use outputs.
- • Limited preview access was restricted to trusted partners following U.S. government pressure, with general availability subsequently rolled out for the GPT-5.6 family.
Summary
OpenAI announced the general availability of its GPT-5.6 family of models, positioned as next-generation frontier intelligence designed to scale across enterprise and professional use cases. The family comprises three tiers: Sol (flagship), Terra (balanced efficiency model), and Luna (most cost-efficient variant), allowing organizations to optimize for capability, speed, or cost depending on workload requirements.
GPT-5.6 Sol established benchmarking leadership across multiple evaluation frameworks. On Agents' Last Exam—a comprehensive assessment of long-running professional workflows spanning 55 fields—Sol achieved 53.6 points, exceeding Claude Fable 5's adaptive reasoning score by 13.1 points. The model demonstrated particular strength in presentation quality outputs, producing more polished and accurate results in presentations, documents, and spreadsheets. On the Artificial Analysis Intelligence Index, a broad measure spanning agentic work, coding, scientific reasoning, and general capabilities, Sol with maximum reasoning came within one point of Fable 5 while completing tasks in 61% less time at approximately half the estimated cost.
Cost efficiency represents a material competitive differentiator. GPT-5.6 Terra and Luna outperform Claude Fable 5 at approximately one-sixteenth the estimated cost, addressing a critical market dynamic for organizations deploying AI at scale. At medium reasoning settings, Sol itself beats Fable 5 by 11.4 points at roughly one-quarter the cost. OpenAI emphasized "stronger performance per dollar"—the ability to deliver comparable results at lower total cost or superior outcomes for equivalent spend.
The company introduced "ultra," a new high-capability mode that coordinates multiple agents across parallel workstreams to accelerate complex task completion. Enhanced computer use and design judgment capabilities position Sol as what OpenAI characterizes as "the most polished collaborator yet," capable of inspecting, refining, and delivering production-ready outputs. The model was optimized to extract greater utility from each input token, addressing both performance and efficiency metrics simultaneously.
Initial deployment followed a limited preview phase restricted to trusted partners, reflecting regulatory considerations related to U.S. government guidance on frontier model access and deployment protocols.
GPT-5.6 Sol established benchmarking leadership across multiple evaluation frameworks. On Agents' Last Exam—a comprehensive assessment of long-running professional workflows spanning 55 fields—Sol achieved 53.6 points, exceeding Claude Fable 5's adaptive reasoning score by 13.1 points. The model demonstrated particular strength in presentation quality outputs, producing more polished and accurate results in presentations, documents, and spreadsheets. On the Artificial Analysis Intelligence Index, a broad measure spanning agentic work, coding, scientific reasoning, and general capabilities, Sol with maximum reasoning came within one point of Fable 5 while completing tasks in 61% less time at approximately half the estimated cost.
Cost efficiency represents a material competitive differentiator. GPT-5.6 Terra and Luna outperform Claude Fable 5 at approximately one-sixteenth the estimated cost, addressing a critical market dynamic for organizations deploying AI at scale. At medium reasoning settings, Sol itself beats Fable 5 by 11.4 points at roughly one-quarter the cost. OpenAI emphasized "stronger performance per dollar"—the ability to deliver comparable results at lower total cost or superior outcomes for equivalent spend.
The company introduced "ultra," a new high-capability mode that coordinates multiple agents across parallel workstreams to accelerate complex task completion. Enhanced computer use and design judgment capabilities position Sol as what OpenAI characterizes as "the most polished collaborator yet," capable of inspecting, refining, and delivering production-ready outputs. The model was optimized to extract greater utility from each input token, addressing both performance and efficiency metrics simultaneously.
Initial deployment followed a limited preview phase restricted to trusted partners, reflecting regulatory considerations related to U.S. government guidance on frontier model access and deployment protocols.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.