GPT-6.1 Sol

Published frontier peer. Closed API gpt-6.1-sol. Index Terminal-Bench 2.1, SWE-bench Verified, and AutomationBench stay empty: the Sep 29, 2026 launch page gives relative deltas only, and absolute chart numbers are not in the page HTML, so those cells are not banked. Relative claims, not Index scores: AutomationBench +2.2 percentage points vs Opus 5.5 medium and +4.8 percentage points vs GPT-6 Sol at the same setting; DeepSWE +6.4 percentage points vs GPT-6 Sol best; OSWorld 2.0 +7 percentage points vs Sol. Terminal-Bench Science 0.1 is named and is not Terminal-Bench 2.1. No SWE-bench Verified line. DeepSWE, OSWorld, and Terminal-Bench Science 0.1 are not Index columns. Not a LocalsOnly desk run. API docs: $2 / $10 per 1M input / output tokens; cached input $0.10; cache writes $2.50; context 1,050,000; max output 128,000; knowledge cutoff Apr 30, 2026; reasoning.effort low, medium (default), high, xhigh, max (no none or minimal). https://openai.com/index/introducing-gpt-6-1-sol https://developers.openai.com/api/docs/models/gpt-6.1-sol