← Back to blog
GPT-6.1 Sol: Pricing, What Changed, and How to Use It

GPT-6.1 Sol: Pricing, What Changed, and How to Use It

Julian Brooks

By Julian Brooks

MyClaw Editorial

GPT-6.1 Sol

Ask GPT-6.1 Sol anything

Use it for difficult coding, deep reasoning, and research tasks that need an expert second opinion.

Chat with GPT-6.1 Sol

AI Takeaway

  • What is GPT-6.1 Sol? OpenAI's successor to GPT-6 Sol, designed for complex coding and professional work at a lower price than Astra.
  • What does it cost? Standard API rates are $2 per million input tokens, $0.10 for cached input, and $10 for output. Long context and processing modes affect the bill.
  • Where can you use it? In the API and supported ChatGPT Work and Codex clients, subject to account access. It is not available in standard Chat.
  • What changes when you upgrade? Check reasoning settings and tool compatibility: none is unsupported, and tool calls require the Responses API.
  • Is it worth upgrading from GPT-6 Sol? It is worth testing if you want stronger results at the same standard input and output rates. Cached input is cheaper, but compatibility changes need attention.

What Is GPT-6.1 Sol, and What Changed?

GPT-6.1 Sol is OpenAI's updated model for coding, computer use, and professional work. Its API identifier is gpt-6.1-sol. OpenAI describes it as approaching Astra's performance at a lower cost, making it a candidate for demanding jobs you run regularly.

Compared with GPT-6 Sol, the clearest specification changes concern cached input and API compatibility:

SpecificationGPT-6 SolGPT-6.1 Sol
Standard input / output per million tokens$2 / $10$2 / $10
Cached input per million tokens$0.20$0.10
Context window1,050,000 tokens1,050,000 tokens
Maximum output128,000 tokens128,000 tokens
none reasoningSupportedUnsupported
Tool calling through Chat CompletionsOnly with none reasoningUnsupported; use Responses

Both accept text and image inputs and produce text output, with no native audio or video support.

The cached-input rate has halved; the standard input and output rates have stayed the same. Any further savings depend on how many tokens and attempts the new model needs to finish your work. OpenAI's capability claims provide a reason to test it, while your own results determine whether to switch.

In an agent workflow, the model supplies reasoning. OpenClaw connects that model to tools, files, memory, and scheduled work. Changing models can therefore affect both answer quality and how those tools operate.

GPT-6.1 Sol Pricing: What a Task Actually Costs

GPT-6.1 Sol is here., Upgraded with stronger agentic coding and computer  use, near-Astra performance, and cached input at a 95% discount to standard  input pricing., GPT-6.1 Sol is built for complex ...

Standard Rates, Caching, and Long Context

As of September 30, 2026, OpenAI lists these Standard API rates for requests with up to 272,000 input tokens:

Token categoryPrice per million tokens
Input$2.00
Cached input$0.10
Cache writes$2.50
Output$10.00

Above that threshold, input and cache rates double, while output rates rise by 50%. These higher rates apply to the full request. Budget for that price change before loading a large document collection.

Fast mode costs twice Standard; Batch and Flex cost half Standard where supported. Keep the processing mode consistent when comparing bills.

A Simple Cost Example

A request using 100,000 uncached input tokens and 10,000 billed output tokens costs $0.30 in Standard token charges: $0.20 for input plus $0.10 for output.

This is an illustrative calculation, excluding cache writes, tool fees, and other surcharges. The 10,000 output tokens must include any billable reasoning tokens; counting only the visible answer would underestimate usage.

For repeated work, use the cached-token count reported in your usage data. Reusing a document does not mean every token qualifies for the cached rate.

An agent may make several requests for one job, so add their costs together, including retries. You can also save on unnecessary input by reducing OpenClaw token overhead: remove irrelevant files from the input and keep recurring context concise. Model choice and context management both affect the bill.

Where Can You Use GPT-6.1 Sol?

You can access GPT-6.1 Sol through OpenAI's API or supported ChatGPT Work and Codex clients. Within ChatGPT, look in Work: the model is not available in standard Chat.

The launch rollout covers Plus, Pro, Business, Enterprise, and Edu. Enterprise and Edu administrators must enable it, while Free and Go are excluded at launch. If it is missing from your model picker, check your plan, client, and workspace settings.

For API use, request gpt-6.1-sol through an account with access. A ChatGPT subscription and API billing are separate arrangements. If you use a third-party provider, confirm its model identifier, endpoint support, and rates instead of assuming OpenAI's direct pricing applies.

If you also need an always-on agent, budget for its runtime. MyClaw hosting plans describe the hosted environment and any included credits; check those allowances alongside your expected model usage.

Sam Altman announces the release of GPT 6.1 Sol, OpenAI's most powerful  model yet. | Yahoo Finance | Facebook

GPT-6.1 Sol vs Astra and Luna: Which Should You Try?

OpenAI positions Astra for its most demanding work, Sol for a balance of capability and cost, and Luna for efficient processing of focused tasks. For Standard requests below the long-context threshold, Sol's $2 input and $10 output rates per million tokens are one-fifth of Astra's $10 and $50. That is a per-token comparison; task costs also depend on usage.

Use the workload to choose your starting point:

  • Try GPT-6.1 Sol for complex recurring work. Examples to evaluate include code changes, document preparation, and tasks that require several tool calls. Its appeal is the balance between capability and cost.
  • Keep Astra in the comparison for your hardest jobs. Difficult debugging or ambiguous analysis may justify a higher rate if it produces a usable result with less correction.
  • Try Luna for focused, high-volume tasks. Extraction, classification, and structured summaries are good candidates when success is easy to define and check.

For a useful comparison, give each model the same source material, tools, and acceptance criteria. Record reasoning settings and compare completion quality, elapsed time, retries, and total billed usage. Hold the task constant rather than changing the prompt to favor one model.

Choose the least expensive option that consistently meets your requirements. Include correction time in that judgment: a model that saves a few cents but requires repeated manual fixes may be a poor fit for daily work.

Try GPT-6.1 Sol on your own work

Ask GPT-6.1 Sol a complex question, coding task, or research prompt right away.

Chat with GPT-6.1 Sol

Before You Switch an OpenClaw Agent

Check Reasoning and Tool Support

GPT-6.1 Sol supports low, medium, high, xhigh, and max API reasoning effort, with medium as the default. If your integration currently sends none or minimal, OpenAI recommends starting with low and comparing results. Work/Codex interface labels may differ from these API values.

Tool calling requires the Responses API. An agent that previously used GPT-6 Sol through Chat Completions with none reasoning needs more than a new model name. Its provider integration must support the required request and tool workflow.

Check your provider's model listing and the runtime's Responses support together. Then test a tool call, its returned result, and the model's next action. A successful text reply alone does not confirm that the agent can complete this sequence.

Test the Whole Workflow

Choose a bounded task: read three supplied files, produce a weekly report, and save it in a specified location. Define the required sections and totals in advance.

Inspect the saved file, compare its figures with the sources, check for missing records, and confirm that it opens correctly. Review failed tool calls and record any manual corrections. Keep the previous configuration available until the new one passes these checks.

For recurring jobs, decide who handles updates, restarts, and backups. The choice between managed and self-hosted OpenClaw determines that operational responsibility alongside your model choice.

Evaluate Your Next Agent Model with MyClaw

A weekly report needs its files, tools, and runtime available when the job starts. MyClaw's managed OpenClaw hosting provides a private, dedicated environment with always-on operation, updates, and backups, reducing the server maintenance you handle.

Chat with GPT-6.1 Sol

GPT-6.1 Sol support must still be confirmed for your provider and workspace. To evaluate it when available:

  1. Start with one recurring job. Prepare your MyClaw OpenClaw workspace around a specific deliverable, such as a weekly report from a defined folder. Supply the files, output format, and criteria for a complete result.
  2. Connect a supported model. Verify GPT-6.1 Sol access and Responses-compatible tool calling before selecting it. If either is unavailable, continue with your current supported model until the integration is ready.
  3. Review a complete run. Compare the artifact, elapsed time, retries, and usage against your current model. Inspect the result before making the configuration your default or scheduling repeated runs.

Start with the next report you already need. If the new setup produces a complete, accurate file at an acceptable cost across repeated runs, you have a practical reason to adopt it.

Put GPT-6.1 Sol to work on a real task

Start a chat with GPT-6.1 Sol and test its reasoning on the work that matters to you.

Chat with GPT-6.1 Sol
GPT-6.1 Sol: Pricing, What Changed, and How to Use It | MyClaw.ai