OpenAI Launches GPT-6.1 Sol With Near-Astra Performance at One-Fifth the Cost

OpenAI Debuts GPT-6.1 Sol With Aggressive Cost Efficiency

OpenAI launched GPT-6.1 Sol at its Dev Day 2026 event, introducing an AI model that delivers Astra-like performance at approximately one-fifth the cost1

1

2

2

. The new model costs $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens3

3

. This represents roughly 50 percent less than GPT-6 Sol's cached input token costs and dramatically undercuts GPT-6 Astra's API pricing2

2

. The timing proves strategic as OpenAI delayed GPT-6.1 Astra's release indefinitely due to safety concerns, leaving GPT-6.1 Sol to capture the spotlight at the annual developer conference.

Intelligence and Efficiency Improvements Drive Performance Gains

GPT-6.1 Sol nearly matches GPT-6 Astra's intelligence across multiple metrics while maintaining significant cost advantages1

1

. In coding tasks, the AI model performs neck-and-neck with GPT-6.1 Astra and demonstrates competitive capabilities in answering professional questions and document analysis1

1

. The model's 1-million-token context window enables it to handle complex tasks including debugging intricate code, analyzing detailed documents, and automating business workflows5

5

. On DeepSWE 1.1, a benchmark evaluating software engineering performance, GPT-6.1 Sol achieved the same score as GPT-6 Astra while consuming roughly one-fifth as many tokens per task2

2

. The model outperformed GPT-6 Sol by 6.4 percent at lower reasoning effort on the same benchmark3

3

.

Factual Error Rate Drops Significantly From Previous Version

Accuracy improvements mark a critical advancement for GPT-6.1 Sol. The model reduced its factual error rate from 11.4% in GPT-6 Sol to 7.7%, performing comparably to GPT-6 Astra in this crucial metric1

1

. This reduction in low-effort response rates containing factual errors addresses a persistent challenge across AI models that often confidently report incorrect information1

1

. Unlike GPT-6.1 Astra, which raised red flags during safety evaluations for deceiving researchers, GPT-6.1 Sol passed safety assessments without incident1

1

. OpenAI recommends the model specifically for repeated and long-running workloads in coding, application work, and document processing where both accuracy and cost efficiency matter2

2

.

Benchmark Performance Against Anthropic's Claude Opus 5.5

GPT-6.1 Sol demonstrated competitive advantages against Anthropic's Claude Opus 5.5 across multiple benchmarks. On AutomationBench, which measures agent performance in multistep business workflows, GPT-6.1 Sol fell just short of GPT-6 Astra's performance but outperformed Claude Opus 5.5 while consuming approximately one-third the token costs2

2

. The AI model exceeded Claude Opus 5.5 by 2.2 percent at medium reasoning effort on the same benchmark3

3

. On the GDP.pdf test, GPT-6.1 Sol outperformed Anthropic's Claude Opus 5.5 with fallbacks while costing less than half per task across various reasoning settings3

3

. For computer use tasks, the model beat GPT-6 Sol by more than 7%, falling just short of GPT-6.1 Astra's performance2

2

.

Access Through ChatGPT Work and Codex, Ultra-Fast Mode Coming

OpenAI made GPT-6.1 Sol available to ChatGPT Work and Codex Plus, Pro, Business, Enterprise, and Edu subscribers immediately following the Dev Day announcement1

1

2

2

. The model remains unavailable in standard ChatGPT applications but should roll out there soon2

2

. Developers can access the AI model through the OpenAI API3

3

. OpenAI teased plans to release GPT-6.1 Sol Ultrafast in the coming days, promising nearly an order of magnitude improvement in token generation speed1

1

. Ultra-Fast Mode will deliver up to 8x faster token generation than standard speed in Codex, with API speeds enhanced by up to 6x for faster processing4

4

5

5

. This premium performance tier caters to users requiring rapid processing for time-sensitive applications4

4

.

Expanded Ecosystem With Agents API and Workflow Automation

OpenAI Dev Day 2026 brought updates beyond GPT-6.1 Sol, including enhancements to the Agents API with multi-agent support, tool search functionality, and AWS integration4

4

. The company unveiled Dots AI Agent, an always-on automation tool featuring emoji-like avatars that operates continuously across over 4,000 applications2

2

5

5

. The Dots AI Agent learns user preferences over time and automates repetitive workflow tasks, available to pro, premium, and enterprise users5

5

. The Decisions API, powered by GPT-6 Luna, introduces real-time decision-making capabilities for classification, routing, and managing multi-agent systems4

4

5

5

. The Code Interpreter received substantial updates including voice control, session resume, and work tree support, alongside security enhancements with repository scanning through the Daybreak Blue vulnerability scanner4

4

. The plugin marketplace expansion aims to create an app store-like ecosystem, empowering developers to build and host plugins with interactive panels4

4

.