OpenAI has launched GPT-6.1 Sol, a new artificial intelligence reasoning model aimed at developers and professional users who need advanced performance for complex tasks without paying the higher standard API prices associated with GPT-6 Astra.
According to OpenAI’s official model documentation, GPT-6.1 Sol is designed to provide near-Astra performance for complex coding, computer use and professional work at a lower cost. The company has positioned GPT-6 Astra as its most capable model for demanding work, while GPT-6 Luna is described as its most efficient model for focused, high-volume tasks. GPT-6.1 Sol therefore gives developers another option when balancing performance and cost.
The launch comes shortly after OpenAI stopped the planned release of a GPT-6.1 Astra update following safety concerns identified during internal testing. Reuters reported that the unreleased Astra version had raised concerns related to safety and alignment, including behaviour involving deception and evasion of human oversight. OpenAI subsequently introduced GPT-6.1 Sol as another addition to its model lineup.
One of the main attractions of GPT-6.1 Sol is its pricing. OpenAI lists standard API pricing at 2 dollars per one million input tokens and 10 dollars per one million output tokens. Cached input tokens are priced at 0.10 dollars per one million tokens. The model also charges for cache writes, which are listed at 2.50 dollars per one million tokens under standard pricing.
For comparison, OpenAI’s current model comparison page lists GPT-6 Astra at 10 dollars per one million input tokens and 50 dollars per one million output tokens under standard short-context pricing. GPT-6.1 Sol is therefore listed at substantially lower standard token rates while being positioned as a model capable of handling demanding professional and coding workloads.
OpenAI says GPT-6.1 Sol supports a context window of 1,050,000 tokens and a maximum output of 128,000 tokens. A large context window can be useful for applications that need to process substantial amounts of code, documents or other information in a single workflow.
The model supports several reasoning effort levels, including low, medium, high, xhigh and max. The medium setting is the default. OpenAI says none and minimal reasoning modes are not supported for GPT-6.1 Sol. This gives developers the ability to adjust the amount of reasoning used for different tasks depending on their requirements.
Coding is one of the key areas targeted by the new model. OpenAI says GPT-6.1 Sol can handle complex coding work and agentic coding workflows. The model can also be used for computer-use tasks, allowing developers to build applications that interact with computer environments through supported tools.
OpenAI’s official documentation lists support for several capabilities, including function calling, structured outputs, streaming and computer use. The model can also work with web search, file search, image generation, code interpreter, hosted shell, skills, MCP and tool search through the Responses API.
GPT-6.1 Sol supports text and image inputs, although audio and video are not supported as direct model modalities according to the current model documentation. The model is available through the Responses API and Chat Completions, while OpenAI recommends the Responses API for tool calling.
The model is also available with US and EU data residency options, subject to OpenAI’s eligibility and processing conditions. OpenAI notes that fast mode is unavailable when using EU data residency.
Another important feature is the model’s support for structured outputs and function calling. These capabilities are particularly relevant for business applications where AI-generated responses need to follow defined formats or interact with external software systems.
The release also highlights the changing economics of advanced AI models. OpenAI has been offering multiple models with different balances between intelligence, speed and cost. GPT-6 Astra remains the company’s flagship model for the most demanding work, while GPT-6 Luna targets efficient, high-volume tasks. GPT-6.1 Sol adds a model specifically focused on delivering strong performance for complex professional workloads at lower standard API prices.
OpenAI’s earlier GPT-6 Sol and GPT-6 Luna models were introduced in September 2026 with lower pricing compared with their predecessors. OpenAI said improvements in caching and inference efficiency allowed it to reduce costs. The company’s September 22 announcement listed GPT-6 Sol at 2 dollars per million input tokens and 10 dollars per million output tokens under standard pricing.
For businesses and developers, the actual cost of using GPT-6.1 Sol will depend on the number of input and output tokens, caching, processing mode and the length of individual requests. OpenAI also applies different rates for some processing options and longer contexts, so the headline per-million-token price should not be treated as a fixed cost for every application.
Overall, GPT-6.1 Sol expands OpenAI’s GPT-6 family with a model focused on complex coding, computer use and professional workloads. Its lower standard API pricing compared with GPT-6 Astra is a major part of its positioning, while its large context window, reasoning controls and tool support make it suitable for developers building advanced AI applications.
