OpenAI: GPT Sol Latest

GPT-6.1 Sol is OpenAI's September 29, 2026 upgrade to GPT-6 Sol in the GPT-6 family. It is positioned below flagship GPT-6 Astra as a cost-efficient reasoning model for complex coding, computer use, and professional workflows, with standard API pricing at $2 per million input tokens and $10 per million output tokens (cached input reads at $0.10 per million). The model supports a 1.05 million-token context window (up to 922K input and 128K output tokens), text and image input, configurable reasoning effort from low through max, and agent tooling via the Responses API.

Relative to GPT-6 Sol, OpenAI reports meaningful gains on agentic engineering and desktop automation benchmarks while keeping token prices unchanged. Public disclosures highlight DeepSWE v1.1 at 75.2% (high reasoning effort), AutomationBench 1.0.6 at 31.7% (medium effort), and OSWorld 2.0 offline partial reward at 71.4% (max effort), approaching Astra at substantially lower cost per task. Independent and cataloged evaluations also place the model competitively on LiveBench, Humanity's Last Exam, and MMMU-Pro at max effort. The GPT-6.1 Sol system card addendum treats cybersecurity risk as Critical and biological/chemical risk as High, applying Astra-aligned safeguards.

GPT-6.1 Sol is available through the OpenAI API as gpt-6.1-sol, in ChatGPT Work and Codex for paid tiers, and is intended for teams that need near-frontier capability on long-horizon coding and tool-using tasks without defaulting to Astra pricing. Parameter count and detailed architecture are not published; weights are proprietary and not downloadable for local deployment.