GPT-6 Sol Explained: Pricing, Benchmarks, GPT-6.1 Sol and How to Choose

GPT-6 Sol is the mid-priced tier of OpenAI's GPT-6 family, released on 22 September 2026 together with the cheaper GPT-6 Luna. It costs $2 per million input tokens and $10 per million output tokens, half the price of GPT-5.6 Sol, and brings much of the flagship GPT-6 Astra's reasoning to everyday workloads. A week later, at DevDay, OpenAI shipped GPT-6.1 Sol at the same price with stronger agentic performance.

That pace is the story. In under three months teams have seen GPT-5.6 Sol, a promotional price cut, GPT-6 Sol and GPT-6.1 Sol. This guide sets out what each model is, what the numbers say, and how to choose without rewriting your product every fortnight. If you are still on GPT-5.6, start with our GPT-5.6 Sol guide, and for the flagship see GPT-6 Astra explained.

The GPT-6 family at a glance

Both GPT-6 Sol and Luna offer a context window of about 1.1 million tokens with multimodal input, according to launch coverage from Computing for Geeks and Vellum.

What changed from GPT-5.6 Sol

• Price halved: $2 / $10 versus GPT-5.6 Sol's promotional $4 / $20. OpenAI's worked example: a workload costing $800 on GPT-5.6 Sol costs $400 on GPT-6 Sol.
• No expiry on the new rates: an OpenAI spokesperson told VentureBeat the prices are not promotional.
• Fewer factual errors: OpenAI says GPT-6 Sol makes roughly half as many factual errors as its predecessor, approaching Astra's reliability.
• Luna got much cheaper: $0.10 / $0.50, down from $0.20 / $1.20, which makes it realistic for classification and routing at scale.

Benchmarks: what the numbers say

Launch benchmarks reported by independent trackers place GPT-6 Sol clearly above Luna and below Astra. Reported figures include an Intelligence Index of 48 and a Coding Agent Index of 57 at maximum effort for Sol, against 37 and 41 for Luna, and 33.2% versus 20.7% on AutomationBench. Treat single numbers with care: they depend on effort settings and harnesses, and your own tasks are the only benchmark that matters.

Against competitors, comparisons from TechRepublic and others show Claude Opus 5.5 ahead on several demanding coding and agent benchmarks, with one high-effort test reporting 57% versus 26% on Terminal-Bench 4.0. GPT-6 Sol's advantages are price, roughly half of Opus 5.5 per token, and the option to switch extended reasoning off for simple jobs. The common recommendation: use the cheaper model for routine implementation, tests and high-volume edits, and the stronger one when the task is hard and must be right first time.

GPT-6.1 Sol: the upgrade one week later

At DevDay on 29 September OpenAI released GPT-6.1 Sol, positioned as near-Astra intelligence for coding, computer use and professional work at one-fifth of Astra's token prices. OpenAI reports that it matches GPT-6 Astra on DeepSWE v1.1, a benchmark of complex software-engineering tasks in real codebases, and that at maximum effort it averaged $5.47 per task on a science benchmark where Astra averaged $23.80, according to The Next Web.

Which tier to use for what

This is the model-routing pattern we recommend in the AI automation guide: cheap models for most steps, strong models for the few steps that decide the outcome. For coding teams, see also AI coding agents for software teams.

How to migrate without breaking production

1. 01Move model names into configuration so a switch is a config change, not a code change.
2. 02Build an evaluation set of 50 to 200 real tasks with known good outputs, including tool calls and edge cases.
3. 03Run old and new models side by side and compare accuracy, latency and cost per completed task, not cost per token.
4. 04Shadow first: send a copy of live traffic to the new model without using its output.
5. 05Roll out gradually by feature or percentage of traffic, with automatic fallback to the old model on errors.
6. 06Re-check prompts and tool schemas; newer models sometimes follow instructions more literally.
7. 07Review monthly; prices and models are still moving quickly.

The same discipline applies to agents: measure outcomes, cap spend per run and keep humans on irreversible actions, as covered in how to measure AI automation ROI and OpenAI Dots explained.

Sources and further reading

• OpenAI: Introducing GPT-6 Sol and Luna
• OpenAI: Introducing GPT-6.1 Sol
• OpenAI: API pricing
• VentureBeat: GPT-6 Sol and Luna cut API costs 50% or more
• The Next Web: GPT-6.1 Sol at a fifth of Astra's price
• TechRepublic: GPT-6 Sol vs Claude Opus 5.5