OpenAI: GPT-6 Sol Pro

GPT-6 Sol Pro is OpenAI's pro reasoning execution mode for the GPT-6 Sol model, exposed in the API as openai/gpt-6-sol-pro. It uses the same weights as GPT-6 Sol but sets reasoning.mode to pro so the system spends more internal reasoning before returning a final answer. OpenAI and OpenRouter position it for hard, high-stakes work—complex coding, multi-step agents, and professional workflows—where extra accuracy is worth higher latency and substantially higher reasoning-token usage than standard Sol requests.

The model ships as a closed, API-only service with a 1.05M-token context window, up to 128K completion tokens, and text plus image inputs with text outputs. List pricing matches GPT-6 Sol at $2 per million input tokens and $10 per million output tokens, with prompt caching discounts; pro mode typically multiplies total cost because each call consumes far more reasoning tokens even though per-token rates are unchanged. It supports tool calling, structured outputs, adjustable reasoning effort, and the broader GPT-6 agent stack used in Codex and ChatGPT Work.

Reported evaluations for the Sol family at max or xhigh effort highlight strong agentic coding and workflow scores, including DeepSWE v1.1 at 68.8%, AutomationBench at 33.2%, LiveBench near 79, and competitive long-horizon agent benchmarks such as Terminal-Bench 2.1. GPT-6 Astra remains OpenAI's top-tier model for the hardest tasks; Sol Pro targets teams that want near-frontier Sol capability with deeper reasoning per request without moving to Astra pricing.