--- url: "https://www.xcademia.com/news/gpt-6-1-sol-openai-brings-near-astra-performance-to-a-lower-cost-model" title: "GPT-6.1 Sol: OpenAI Brings Near-Astra Performance to a Lower-Cost Model" description: "OpenAI introduces GPT-6.1 Sol with near-Astra performance for coding, AI agents, computer use and professional workflows at a lower cost." publishedAt: "2026-09-30T10:48:38.473+00:00" updatedAt: "2026-10-10T02:43:53.465298+00:00" type: news category: "ai-ml" source_name: OpenAI source_url: "https://openai.com/index/introducing-gpt-6-1-sol/ " tags: - "#GPT61Sol" - "#OpenAI" - "#ArtificialIntelligence" - "#AIAgents" - "#AIModels" - "#CodingAI" - "#GenerativeAI" - "#MachineLearning" --- # GPT-6.1 Sol: OpenAI Brings Near-Astra Performance to a Lower-Cost Model > OpenAI introduces GPT-6.1 Sol, an upgraded model focused on coding, professional work, computer use and scientific research, with near-Astra performance at a lower API cost. Source: **OpenAI** · 30 September 2026 **O**penAI has introduced **GPT-6.1 Sol**, an upgrade to GPT-6 Sol designed to deliver stronger performance on complex coding, professional workflows, computer-use tasks and scientific research. The company says GPT-6.1 Sol approaches the capabilities of its more powerful GPT-6 Astra model on several evaluations while being offered at a substantially lower cost. OpenAI is positioning the model as an option for developers and organizations that need capable AI agents without using the highest-cost model for every task. GPT-6.1 Sol is currently available through the OpenAI API and in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. OpenAI says it is not yet available in regular ChatGPT. ## What is GPT-6.1 Sol? GPT-6.1 Sol is the latest update to the Sol model family. According to OpenAI, the model focuses on complex work that requires reasoning, tool use and multi-step execution. The company describes the model as offering near-Astra intelligence for agentic coding, computer use and professional work at one-fifth of Astra's standard input and output token prices. This distinction is important because the model is not simply being positioned as a general chatbot upgrade. OpenAI is emphasizing workflows where an AI system needs to work through several steps, understand complex information, interact with tools or applications, and produce a useful result. The model is also designed to make repeated AI workflows more economical through lower cached-input pricing. ![gpt-6-family](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1790761730149-gpt-6-family.webp) ## Professional workflows and complex documents GPT-6.1 Sol is also designed for professional tasks involving documents and multi-step business processes. OpenAI reports results from **GDP.pdf**, an evaluation that tests how accurately models answer professional questions using complex PDF documents. These documents can contain tables, charts, diagrams and detailed information. The evaluation covers professional domains including finance, healthcare and legal work. OpenAI says GPT-6.1 Sol scored higher than Opus 5.5 with fallbacks at less than half the cost per task across the tested reasoning settings. The company also says the model approaches GPT-6 Astra's performance at approximately one-fifth of the cost per task. The model was also tested with **AutomationBench**, an evaluation focused on whether AI agents can complete multi-step business workflows. OpenAI reports that GPT-6.1 Sol scored 2.2 percentage points above Opus 5.5 at medium reasoning effort and was approximately one-third the cost in that comparison. The model also improved by 4.8 percentage points over GPT-6 Sol under the same setting. AutomationBench tests workflows across areas including sales, marketing, operations, support, finance and human resources. These results point toward a broader use case for AI agents: not just answering questions, but completing sequences of related tasks using multiple tools. ![gpt-6-workflow](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1790761914437-gpt-6-workflows.webp) ## Computer use gets a major upgrade Another important area for GPT-6.1 Sol is computer use. Computer-use models are designed to interact with software environments rather than simply returning text. This makes them relevant to tasks where an AI agent needs to navigate applications, perform actions and complete workflows. OpenAI evaluated GPT-6.1 Sol using **OSWorld 2.0**, which measures AI agents performing demanding computer-use workflows. According to OpenAI, GPT-6.1 Sol outperformed GPT-6 Sol by seven percentage points on the offline set at maximum reasoning effort while costing less than half as much per task. The company also says GPT-6.1 Sol came within 2.1 percentage points of GPT-6 Astra's score while costing approximately one-seventh as much per task. These results are based on the specific OSWorld evaluation environment. They should not be interpreted as a guarantee that an AI agent will perform equally well across every desktop application or business environment. For organizations experimenting with computer-use agents, however, the results show why cost-efficient models could become important. Computer-use workflows may involve many individual steps, making the cost of each task relevant when an agent operates at scale. ## Scientific research capabilities GPT-6.1 Sol also targets scientific workflows. OpenAI evaluated the model using **Terminal-Bench Science 0.1**, which tests scientific tasks involving areas such as data analysis, simulation and theorem proving. The company says GPT-6.1 Sol more than doubled GPT-6 Sol's score at maximum reasoning effort while costing less than half as much per task. OpenAI reports an average cost of $5.47 per task for GPT-6.1 Sol at maximum effort in this evaluation, compared with $23.21 for Opus 5.5 and $23.80 for Astra. OpenAI also notes an important limitation: GPT-6 Astra achieved the highest score among the models tested, with a score of 68.1%, and the company recommends Astra for the most difficult scientific research tasks. This distinction helps explain the intended position of GPT-6.1 Sol. It is designed to bring strong performance to more workflows at a lower cost, rather than replacing the highest-capability model in every situation. ![gpt-6-work](https://0a515t3ure77wbvx.public.blob.vercel-storage.com/articles/1790762010286-gpt-6-work.webp) ## Factual accuracy also improves OpenAI says GPT-6.1 Sol improves factual accuracy compared with GPT-6 Sol on the company's difficult factuality evaluation. At low reasoning effort, the share of responses containing a factual error decreased from 11.4% with GPT-6 Sol to 7.7% with GPT-6.1 Sol. OpenAI describes this as approximately a 32% reduction. However, the company provides an important qualification. The evaluation uses de-identified ChatGPT conversations where users had previously flagged factual errors. These conversations were deliberately selected because they were difficult and are not representative of typical usage. That means the reported error rate should be viewed as an evaluation measurement rather than a direct estimate of how frequently GPT-6.1 Sol will produce errors in everyday use. For developers and businesses, this distinction matters. Improvements in benchmarked factuality can be useful, but production systems still need appropriate validation, monitoring and human oversight, especially when outputs influence important decisions. ## Safety and alignment OpenAI also reports improvements in its alignment evaluations for GPT-6.1 Sol. The company says the model is more transparent about limitations and more reliable at following user intent and safety constraints compared with GPT-6 Sol. The evaluation included situations involving broken search tools, explicit restrictions and unauthorized outcomes during agentic tasks. OpenAI reports that GPT-6.1 Sol showed lower failure rates than GPT-6 Sol in several of these areas and observed no attempts to bypass an automated safety reviewer, matching GPT-6 Astra and GPT-6 Sol. OpenAI's deployment safety documentation says GPT-6.1 Sol is classified as **Critical for cybersecurity capability** and **High for biological and chemical capability** under its Preparedness Framework. The model uses the same safeguards stack as GPT-6 Astra, according to the system card addendum. The company also emphasizes that these safety evaluations intentionally create challenging conditions and should not be interpreted as measurements of typical user interactions. ## Pricing and API access Cost is one of the central parts of the GPT-6.1 Sol announcement. OpenAI lists the standard API pricing as: - **Input:** $2 per million tokens - **Cached input:** $0.10 per million tokens - **Output:** $10 per million tokens The cached-input price is particularly notable for applications that repeatedly reuse the same context across requests. For example, an AI coding agent may repeatedly work with the same project context, documentation or instructions. Lower cached-input costs can make these repeated interactions more economical. OpenAI says GPT-6.1 Sol is priced at one-fifth of GPT-6 Astra's standard input and output token prices. The model is available through the API under the model ID **gpt-6.1-sol**. ## Availability in ChatGPT, Codex and the API GPT-6.1 Sol is available to **Plus, Pro, Business, Enterprise and Edu users** in ChatGPT Work and Codex, according to OpenAI. It is currently **not available in regular ChatGPT**. Developers can access the model through the OpenAI API using the `gpt-6.1-sol` model identifier. OpenAI also announced plans to offer an **Ultrafast** version in the coming days, with up to eight times faster token generation compared with standard speed in Codex. Additional details were not disclosed in the announcement about the exact rollout schedule beyond the stated coming-days timeframe. ## What GPT-6.1 Sol means for developers The most important aspect of GPT-6.1 Sol is the combination of capability and cost. AI agents often perform multiple steps instead of generating one response. An agent may inspect files, reason about a problem, call tools, interact with applications and then verify its result. As the number of AI-driven workflows grows, the economics of each task become increasingly important. GPT-6.1 Sol is designed around that problem. OpenAI is offering a model that it says approaches GPT-6 Astra on several demanding evaluations while costing substantially less. For software teams, this could make more complex agentic workflows practical to run at higher volumes. However, benchmark results are not substitutes for testing. Organizations considering GPT-6.1 Sol should evaluate it against their own workloads, data, tools, reliability requirements and human-review processes. ## GPT-6.1 Sol compared with GPT-6 Astra The models serve different positions within OpenAI's lineup. GPT-6 Astra remains OpenAI's highest-capability model for the most demanding workloads. GPT-6.1 Sol is positioned as a lower-cost option that approaches Astra on several evaluated tasks. This creates a model-selection tradeoff rather than a simple replacement. For a workload where maximum capability is essential, Astra may remain appropriate. For workflows where cost and repeated agent execution matter more, GPT-6.1 Sol provides another option. The right choice depends on the requirements of the specific application. ## The broader shift toward cost-efficient AI agents The GPT-6.1 Sol announcement highlights a broader industry shift toward making advanced AI useful not only for individual prompts, but also for repeated and multi-step workflows. As AI moves into coding agents, business automation, computer interaction and research workflows, the cost of running those systems becomes an important engineering consideration. The development also reflects growing demand for models that can combine reasoning with tool use while remaining economical enough for repeated execution. For enterprises, this could mean more experimentation with AI agents across internal workflows, provided organizations establish appropriate security, access controls, monitoring and human oversight. ## Final takeaway GPT-6.1 Sol represents an incremental but significant update within OpenAI's GPT-6 model family. The model focuses on complex coding, professional document work, computer use and scientific workflows. OpenAI's published evaluations show improvements over GPT-6 Sol across these areas, while the company positions GPT-6.1 Sol as a lower-cost alternative to GPT-6 Astra. The model is already available through the API and through ChatGPT Work and Codex for supported paid, business and education plans. It is not yet available in regular ChatGPT. For developers, the most relevant question is likely not simply whether GPT-6.1 Sol is more capable, but whether its combination of performance, reasoning and pricing fits the specific workflows they are building. ## Original source https://openai.com/index/introducing-gpt-6-1-sol/ ## Tags `#GPT61Sol` · `#OpenAI` · `#ArtificialIntelligence` · `#AIAgents` · `#AIModels` · `#CodingAI` · `#GenerativeAI` · `#MachineLearning` --- ## About this content This Markdown news article is the citation-grade twin of [GPT-6.1 Sol: OpenAI Brings Near-Astra Performance to a Lower-Cost Model](https://www.xcademia.com/news/gpt-6-1-sol-openai-brings-near-astra-performance-to-a-lower-cost-model). It is published by **Xcademia** (UK Companies House 12322710) and is available for AI search engines and large language models to index, summarise, and cite. When citing or quoting, please attribute *Xcademia* and link back to the source URL above. - Source: https://www.xcademia.com/news/gpt-6-1-sol-openai-brings-near-astra-performance-to-a-lower-cost-model - Publisher: Xcademia, https://www.xcademia.com - Catalogue index: https://www.xcademia.com/llms-full.txt