OpenAI GPT-6.1 Sol: Near-Astra Intelligence at a Much Lower Cost

OpenAI has introduced GPT-6.1 Sol, a major upgrade to its GPT-6 Sol model. The new model is designed to deliver performance close to OpenAI’s more powerful GPT-6 Astra on agentic coding, computer use and professional work, while costing significantly less. OpenAI says GPT-6.1 Sol is aimed at developers and businesses that need capable AI agents without paying the full price of its flagship model.

The timing is also interesting. OpenAI introduced GPT-6.1 Sol just as the company was expanding its focus on AI agents, coding systems and AI that can complete multi-step work. Instead of making Sol simply a faster chatbot, OpenAI has focused this version on practical work where an AI model needs to reason, use tools, interact with computers and complete longer tasks.

What Is GPT-6.1 Sol?

GPT-6.1 Sol is the latest model in OpenAI’s GPT-6 series. It sits below GPT-6 Astra in the model lineup, but OpenAI says it can approach Astra’s intelligence on several important tasks at a fraction of the cost. The company specifically highlights agentic coding, computer use and professional work.

This makes Sol different from a model designed mainly for casual conversations. It is intended for workloads where the model may need to perform several steps, work with large amounts of information, interact with software and continue working toward a goal.

For developers, that can mean coding agents that can work through repositories and solve complicated programming tasks. For businesses, it can mean automating document-heavy workflows and repetitive computer-based processes.

GPT-6.1 Sol Is Built for AI Agents

One of the biggest themes behind GPT-6.1 Sol is agentic AI.

An AI agent does more than answer a question. It can receive a goal, decide what needs to happen, use tools, inspect the result and continue working. That makes model quality important, but cost is also important because agents can make many model calls during one task.

OpenAI’s lower pricing for Sol is therefore a major part of the product. A developer can potentially run more iterations without the cost becoming as restrictive as it would with a flagship model.

OpenAI says GPT-6.1 Sol nearly matches GPT-6 Astra’s intelligence in agentic coding, computer use and professional work while costing one-fifth of Astra’s standard input and output token prices.

GPT-6.1 Sol Pricing

The API pricing is one of the biggest reasons developers may pay attention to GPT-6.1 Sol.

OpenAI lists the model at:

• $2 per million input tokens
• $10 per million output tokens
• $0.10 per million cached input tokens

The cached input price is particularly interesting. OpenAI says it is 95% lower than standard input pricing and half the cached-input price of GPT-6 Sol.

Caching matters for agents because they may repeatedly work with the same instructions, files or context. Instead of paying the full input price every time, developers can reuse cached information at a much lower cost.

This could make GPT-6.1 Sol attractive for applications that need repeated model calls.

Coding Performance

Coding is one of the main areas where GPT-6.1 Sol is expected to compete with higher-end models.

OpenAI describes the model as an upgrade designed for agentic coding rather than simply traditional code generation. The goal is to let the model work through larger software tasks and operate with development tools.

The model also shows substantial improvements over GPT-6 Sol on computer-use evaluations. OpenAI reports that on OSWorld 2.0, which evaluates agents completing demanding computer workflows, GPT-6.1 Sol improved by seven percentage points over GPT-6 Sol at maximum reasoning effort. It also came within 2.1 percentage points of GPT-6 Astra while costing much less per task.

That difference is important because real-world coding increasingly involves more than writing a few lines of code. An AI coding agent may need to open files, inspect an application, understand errors, make changes and test the result.

Computer Use Is a Major Focus

GPT-6.1 Sol also improves OpenAI’s computer-use capabilities.

Computer-use models are designed to interact with applications in ways that resemble human computer operation. Instead of simply returning instructions, an agent can potentially navigate interfaces and perform a sequence of actions.

This opens the door to automation for repetitive office work, research, software testing and other digital workflows.

For example, an AI agent could potentially move through a business application, collect information, compare records and prepare a result. The model still needs appropriate permissions and monitoring, but the technology is moving toward AI systems that can actually operate software rather than just tell users how to operate it.

Professional Work

OpenAI also positions GPT-6.1 Sol for professional work.

That includes tasks involving documents, research, analysis, coding and business workflows. The lower price makes it particularly interesting for companies that want to integrate AI into everyday operations.

A company does not necessarily need the most powerful model for every task. Some jobs require maximum reasoning capability, while others simply need a strong model that is fast and inexpensive enough to run repeatedly.

That is where Sol fits.

GPT-6.1 Sol vs GPT-6 Astra

The easiest way to understand the difference is to think of Astra as the maximum-performance option and Sol as the efficiency option.

GPT-6 Astra remains OpenAI’s most powerful model, while GPT-6.1 Sol is designed to deliver similar capability on several important workloads at a much lower price. OpenAI’s own description says Sol nearly matches Astra on agentic coding, computer use and professional work.

That does not mean the two models are identical.

For the most difficult reasoning-heavy projects, Astra may still be the better choice. But if an application needs thousands or millions of model interactions, Sol’s pricing can make a major difference.

Safety and Security

The increased capabilities also bring increased safety concerns.

OpenAI’s safety documentation classifies GPT-6.1 Sol as Critical capability in cybersecurity and High capability in biological and chemical domains under its Preparedness Framework. OpenAI says Sol uses the same safeguards stack as GPT-6 Astra.

That classification is significant. It shows that OpenAI is not treating Sol as simply a cheaper version of an ordinary chatbot. The company considers its capabilities strong enough to require substantial safeguards.

OpenAI’s system card also evaluates issues involving model monitoring, cybersecurity and potentially problematic behavior during complex tasks.

What GPT-6.1 Sol Means for Developers

For developers, GPT-6.1 Sol could be one of the more useful models in the GPT-6 family because of its balance between capability and price.

The biggest opportunity is not necessarily replacing every existing model. Instead, developers can choose the model based on the job.

A complicated task that requires maximum intelligence can use Astra. More routine agentic coding, document processing, computer-use tasks and repeated workflows can use Sol.

That type of model selection can reduce the overall cost of running an AI application.

What About Graphic Designers?

GPT-6.1 Sol is not a dedicated graphic design model, but designers can still benefit from it.

Graphic design workflows now include research, content planning, client communication, creative briefs, website work, presentations and repetitive digital tasks. An AI agent capable of computer use could eventually automate some of these repetitive activities.

For example, a designer could use an AI system to organize project information, prepare content, analyze references or handle repetitive software-based tasks while spending more time on actual creative decisions.

The important point is that Sol is more useful as a workflow and productivity model than as a direct replacement for image-generation or design software.

Final Thoughts

GPT-6.1 Sol is an interesting move from OpenAI because it focuses on something developers actually care about: how much useful work an AI model can perform for the money.

The model does not try to replace GPT-6 Astra at the top of the lineup. Instead, it brings much of Astra’s capability closer to a price point that makes large-scale AI agents more practical.

Its strongest areas include agentic coding, computer use and professional work. Its $2 per million input token and $10 per million output token pricing also makes it considerably more affordable than Astra.

The bigger story may be what happens when developers start building around it. If GPT-6.1 Sol can consistently deliver near-frontier performance while keeping agent costs under control, it could become one of OpenAI’s most practical models for real-world AI applications.

For users, the message is simple: GPT-6.1 Sol is not OpenAI’s biggest model, but it may become one of its most useful.