OpenAI’s GPT-5.6 Sol has achieved significant recognition by securing the second position in the Artificial Analysis Intelligence Index, just one point shy of Claude Fable 5, while costing only one third as much. Additionally, GPT-5.6 Sol leads the Artificial Analysis Coding Agent Index with a score of 80, demonstrating strong performance across multiple evaluations. This new model family establishes an innovative pricing structure with cache-write pricing for the first time, and it sets a new efficiency frontier, optimizing intelligence relative to both cost per task and output tokens used.

Grok: Grok is xAI’s large language model designed for general reasoning and tool use. It participates in coding and agentic evaluations alongside other frontier systems. Grok ties for top performance in select sub-benchmarks within the reported coding agent index.
Luna: Luna is the most cost-efficient variant in OpenAI’s GPT-5.6 model family. It achieves solid intelligence scores at significantly reduced per-task costs compared with Sol and Terra. Luna helps expand the range of options along the intelligence versus cost curve.
Terra: Terra is a mid-tier variant in OpenAI’s GPT-5.6 model family. It provides a balance of intelligence and lower task costs than higher variants in the same series. Terra contributes to the family’s overall push of the intelligence-cost Pareto frontier.
OpenAI: OpenAI develops and deploys frontier artificial intelligence models and related systems. The company recently launched the GPT-5.6 model family, including specialized variants for different performance and cost profiles. These releases position OpenAI models competitively against other leading AI systems in independent evaluations.
GPT-5.6 Sol: GPT-5.6 Sol is a high-performance variant in OpenAI’s latest GPT-5.6 model family. It delivers intelligence levels close to top competitors while emphasizing efficiency in reasoning and output tokens. The model also leads specialized coding agent benchmarks when paired with agentic harnesses.
Claude Fable: Claude Fable refers to Anthropic’s frontier model in the Claude series. It leads or places near the top in several independent intelligence and knowledge-work benchmarks. The model serves as a primary point of comparison for cost and performance against OpenAI’s new releases.

Pricing Innovation: GPT-5.6 introduces cache-write pricing at OpenAI for the first time, aligning with practices from other frontier providers.
Efficiency Frontier: The GPT-5.6 family establishes a new Pareto frontier for intelligence relative to both cost per task and output tokens used.
Benchmark Leadership: GPT-5.6 Sol leads the new Artificial Analysis Coding Agent Index across multiple evaluations when used with the Codex harness.