GPT-6 Sol launch: benchmarks, pricing, and when to pin it

What shipped on 22 September

On 22 September 2026 Microsoft added GPT-6 Sol and GPT-6 Luna to generally available Foundry deployments, in GPT-6 Astra, Sol, and Luna for production AI agents. Sol is the cost-efficient high-end row under Astra. Artificial Analysis the same day said the list price is about half of GPT-5.6 Sol: $2 input and $10 output per million tokens, with a 90% cache-read discount and a 25% cache-write premium.

AshnaAI lists the catalog id `gpt-6-sol` on the same Azure Responses path as `gpt-6-astra`. The context window on that row is 1.05 million tokens. Pin it in chat or pass the id through the OpenAI-compatible API.

Benchmarks vs GPT-5.6 Sol and Luna

Read Sol’s column as a cost cut with a small coding gain, not as a new flagship. Coding Agent Index moves from 55 to 57. Terminal-Bench 4.0 in the Codex harness moves from 37% to 43%. SWE-Atlas-QnA moves from 54% to 58%. Hallucination rate on AA-Omniscience falls from 92% to 60%, partly because Sol answers fewer questions (83% attempted versus 99%) and accuracy slips from 59% to 54%. AutomationBench-AA moves from 60% to 62%. Artificial Analysis also says Sol drops about 100 Elo on GDPval-AA v2.1. The Intelligence Index cost per task is $1.06 versus $1.99 for GPT-5.6 Sol.

The table is Artificial Analysis at max effort, dated 22 September 2026. It is an independent scoreboard, not OpenAI’s launch table. GPT-6 Astra still leads OpenAI’s own 3 September computer-use and science cells; those Astra numbers are on the GPT-6 Astra launch post.

What the token bill looks like

Sol at $2 / $10 is the row to pin when GPT-5.6 Sol was the default and the job does not need Astra. Astra remains $10 / $50. Luna is $0.10 / $0.50. Cached input is 10% of the uncached input rate. Cache writes are 1.25× uncached input. On Ashna, reasoning tokens for this family are billed at 1.5× the short-context output rate ($15 per million for Sol).

Long context is a separate Azure rate

Microsoft’s long-context column for Sol is $4 input, $0.40 cached input, $5 cache write, and $15 output. Ashna’s billed sheet stays on the short-context rates. Do not assume a prompt over a few hundred thousand tokens is still $2 / $10 at the provider.

When to pin Sol instead of Astra or Luna

Pin GPT-6 Astra when the job is computer use, science, or the cells Astra published on 3 September. Pin GPT-6 Luna or GLM-5.3-Flash when the job is high volume and a two-point coding-index gap is not worth $10 output. Pin Sol for the middle: agent and coding work that used to sit on GPT-5.6 Sol.

Everyday mixed chat can stay on Ashna-X1, which routes the task instead of locking one id. Cost ladder: Cheapest coding AI vs Claude Opus 5 and GPT-5.6 Sol.

Try gpt-6-sol now

New accounts start at app.ashna.ai/signup. Open GPT-6 Sol in AshnaAI chat. Keep GPT-6 Luna and GPT-6 Astra on the same account so the pin is a switch, not a new vendor.

Product API traffic uses Account → API and the AshnaAI API docs. The model id is `gpt-6-sol`.