GPT-6.1 Sol (OpenAI)

OpenAI released GPT-6.1 Sol at its DevDay event on September 29, 2026, after announcing it would not release the GPT-6.1 Astra variant publicly due to safety concerns. The new model is positioned as a less powerful but safer alternative, reportedly approaching Astra-level performance on coding and intelligence tasks. This continues OpenAI's 2026 naming pattern across its Sol, Terra, and Luna tiered model families. According to the report, GPT-6.1 Sol surpasses the prior GPT-6 Sol on complex tasks like programming, debugging, and multi-step business workflows while reducing error rates. The excerpt states OpenAI claims a token price of $0.10 per million tokens, described as roughly half of GPT-6 Sol's input cost. Astra remains OpenAI's most capable model; 6.1 Sol is positioned as a more accessible counterpart.

Kingy AI's September 29 update confirms GPT-6.1 Sol as the API model name gpt-6.1-sol, with standard rates of $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens. Availability is limited to eligible Plus, Pro, Business, Enterprise, and Edu users in Work and Codex, with Chat rollout not yet enabled. The article's main body documents the September 22 GPT-6 Sol and Luna launch, including shared family specs: 1.05M-token context window, 128K max output tokens, text and image input, and core agent tools including function calling, web search, file search, and computer use. Knowledge cutoffs are April 20, 2026 for Sol and May 18, 2026 for Luna. These figures corroborate the GPT-6.1 Sol lineage context rather than serving as fresh GPT-6.1 Sol benchmarks.

OpenAI introduced GPT-6.1 Sol as an upgrade to GPT-6 Sol, claiming it nearly matches GPT-6 Astra on agentic coding, computer use, and professional work. The model was evaluated on software engineering, professional document understanding, business automation, and computer-use tasks. Improvements span writing and debugging code, document understanding, and executing multi-step business workflows. On DeepSWE v1.1, the model matches GPT-6 Astra at roughly one-fifth the cost and scores 6.4 percentage points higher than GPT-6 Sol. On AutomationBench, it scores 2.2 points above Opus 5.5 at about one-third the cost, and 4.8 points above GPT-6 Sol. On Terminal-Bench Science 0.1, it more than doubles GPT-6 Sol's score while costing less than half per task at maximum reasoning effort.

BenchLM's aggregated page rates GPT-6.1 Sol at capability 77.6/100, ranking it 9th among 211 tracked models. Pricing is confirmed at $2 per million input and $10 per million output, with cached input at $0.10. Context window is 1.05 million tokens, well above the field median of 256,000, while independent runtime speed has not been measured. Category-level ranking shows GPT-6.1 Sol at rank 11 of 119 in Agentic (92nd percentile), rank 22 of 144 in Coding (85th percentile), and rank 5 of 170 in Knowledge (98th percentile). Reasoning, Multimodal, Multilingual, and Instruction Following categories carry no measured benchmarks. The page tracks 11 verified benchmark rows with source-linked evidence as of September 30, 2026.

OpenAI released GPT-6.1 Sol on September 29, 2026, a new model delivering performance close to the GPT-6 Astra flagship at roughly one-fifth the price. API pricing is $2 per million input tokens and $10 per million output tokens, with cached input at $0.10 per million tokens, a 95% reduction from standard input pricing. This positions GPT-6.1 Sol as a cost-down mid-range option in OpenAI's lineup. The model matched GPT-6 Astra on the DeepSWE v1.1 software engineering benchmark and beat GPT-6 Sol by 6.4 percentage points on the same evaluation at lower reasoning effort. On GDP.pdf, GPT-6.1 Sol outperformed Opus 5.5 with fallbacks at less than half the cost per task. Factual errors dropped from 11.4% to 7.7% at low reasoning effort compared to GPT-6 Sol, a 32% reduction on deliberately difficult prompts.