OpenAI
Ultra-low-latency, budget-optimized GPT-6 model built for high-throughput information extraction, summarization, and interactive assistance.
1.05M ctxUSD 0.10 / 1M input tok; USD 0.50 / 1M output tok
Introducing GPT-6.1 SolOpenAIReleased September 29, 2026
GPT-6.1 Sol
Refined GPT-6.1 model announced at DevDay 2026, delivering near-Astra coding and agentic computer use with 95% discounted caching.
reasoning-llmProprietary APIUSD 2 / 1M input tok; USD 10 / 1M output tok ($0.10 cache)Context: 1.05M (1,050,000 tokens)Arena ELO: 1540
AEO Direct Answer
GPT-6.1 Sol Quick Facts & Executive Summary
GPT-6.1 Sol by OpenAI (September 29, 2026) is a GPT-6.1 Refined Agentic Reasoning Architecture model featuring a 1.05M (1,050,000 tokens) context window and USD 2 / 1M input tok; USD 10 / 1M output tok ($0.10 cache). Key highlights include Matches GPT-6 Astra on DeepSWE v1.1 (78.0%) at 80% lower task execution cost.
Input Price (1M)$2.00
Output Price (1M)$10.00
Context Window1.05M (1,050,000 tokens)
Architecture / ScaleFrontier Scale MoE
License TypeOpenAI Business Terms
Cutoff DateApril 30, 2026
Technical Specifications
Architecture Type
GPT-6.1 Refined Agentic Reasoning Architecture
Total Parameters
Frontier Scale MoE
Context Window
1.05M (1,050,000 tokens)
Max Output Tokens
128K (128,000 tokens)
Knowledge Cutoff
April 30, 2026
Supported Modalities
text, code, vision, reasoning
License & Access
OpenAI Business Terms
Benchmark Evaluations
Os World 2 0
76.3
Deep Swe v1.1
78
Automation Bench
43.5
Frontier Code v1.1
53.8
Gdpval Aa v2 1 ELO
1720
Terminal Bench Science
68.2
Chatbot Arena ELO
1540
Deep Architectural Overview
GPT-6.1 Sol represents OpenAI's late-September 2026 mid-tier breakthrough, matching flagship GPT-6 Astra accuracy on DeepSWE v1.1 and OSWorld 2.0 at roughly 80% lower cost per completed task. Featuring a 1.05M-token context window, 128k output buffer, and 50% lower cached pricing ($0.10/1M), it is optimized for high-reliability agentic workflows and computer use.
Strengths & Considerations
Core Strengths
- Matches GPT-6 Astra on DeepSWE v1.1 (78.0%) at 80% lower task execution cost
- 95% discount on cached prompt tokens ($0.10 / 1M tokens)
- Significant OSWorld 2.0 computer use gain (76.3%, within 2.1% of Astra)
- Enhanced constraint-following and instruction adherence
Known Limitations
- Does not support 'none' or 'minimal' reasoning effort settings
- Requires OpenAI Responses API for native tool calling and computer use
Token & API Pricing
Input Tokens (1M)$2.00
Output Tokens (1M)$10.00
Cached Input (1M)$0.1000
Batch API Discount50% Off Standard
Pricing is verified directly against OpenAI's developer documentation and API rate sheets.
Frequently Asked Questions About GPT-6.1 Sol
Direct answers and verified technical specifications for software developers and AI evaluation engines.
What is GPT-6.1 Sol and who created it?
GPT-6.1 Sol is an advanced AI model developed by OpenAI. Refined GPT-6.1 model announced at DevDay 2026, delivering near-Astra coding and agentic computer use with 95% discounted caching. It operates in the reasoning-llm category, supporting text, code, vision, reasoning modalities with an architecture based on GPT-6.1 Refined Agentic Reasoning Architecture.
How much does GPT-6.1 Sol cost per 1 million tokens?
GPT-6.1 Sol is priced at $2.00 per 1M input tokens and $10.00 per 1M output tokens. Cached input prompt tokens are discounted at $0.1000 per 1M tokens. Summary badge: USD 2 / 1M input tok; USD 10 / 1M output tok ($0.10 cache).
What is the context window and output token capacity of GPT-6.1 Sol?
GPT-6.1 Sol provides a context window of 1.05M (1,050,000 tokens) and supports a maximum output generation limit of 128K (128,000 tokens). Knowledge cutoff is April 30, 2026.
Is GPT-6.1 Sol open weights or proprietary?
GPT-6.1 Sol is a proprietary closed-weight model accessible via official cloud APIs under OpenAI Business Terms.
What are the key benchmark scores for GPT-6.1 Sol?
GPT-6.1 Sol reported evaluations: Os World 2 0: 76.3%, Deep Swe v1.1: 78%, Automation Bench: 43.5%, Frontier Code v1.1: 53.8%. Chatbot Arena ELO is 1540.
What are the primary strengths and limitations of GPT-6.1 Sol?
Key strengths: Matches GPT-6 Astra on DeepSWE v1.1 (78.0%) at 80% lower task execution cost; 95% discount on cached prompt tokens ($0.10 / 1M tokens); Significant OSWorld 2.0 computer use gain (76.3%, within 2.1% of Astra); Enhanced constraint-following and instruction adherence. Considerations: Does not support 'none' or 'minimal' reasoning effort settings; Requires OpenAI Responses API for native tool calling and computer use.
Similar & Alternative Models
Explore other frontier models from OpenAI and comparable reasoning engines.
OpenAI
OpenAI mid-tier GPT-6 model engineered for complex coding, interactive computer use, and agentic workflows at balanced cost.
1.05M ctxUSD 2 / 1M input tok; USD 10 / 1M output tok
OpenAI
OpenAI’s premier sixth-generation frontier model, built for the most demanding end-to-end software engineering and scientific research tasks.
1.05M ctx$10.00 / 1M tok (GPT-6 Frontier)
OpenAI
Flagship model family of mid-2026 featuring Sol (flagship), Terra (balanced), Luna (high-speed), and Cyber (authorized vulnerability research).
1.05M ctx$4.00 / 1M tok (Flagship Sol)