# GPT-6 API Update: How To Trial Sol Vs Luna For Drafting And Workflows Source: https://in-store.ca/blog/gpt-6-models-in-the-api Published: 2026-09-24 Reviewed by: Andrew Liashok, Client Service Director, IN-STORE (https://in-store.ca/team/meet-andrew) Publisher: IN-STORE, Toronto, Canada > GPT-6 Sol and Luna are in the API. A practical task split and rollout checklist using the Responses API for content and workflows. ### A Practical Task Split For Service Businesses Using The Responses API For Drafting And Workflows If you run a service business, you rarely have time to experiment slowly. You need a clean way to test new capabilities inside real work, without letting half-finished drafts or fragile automations leak into customer facing channels. OpenAI has added GPT-6 Sol and GPT-6 Luna to the API. Both models accept text and image inputs and generate text via the Responses API and the Chat Completions API. For teams building AI-assisted content and operational workflows, the immediate question becomes practical: which tasks should you trial on Sol versus Luna, and how do you roll those trials out with a safety review and approvals path. Below is a task based split you can use, plus a rollout checklist designed for lean teams that want consistent output and fewer surprises. ## What Changed In The GPT-6 API OpenAI added GPT-6 Sol and GPT-6 Luna to the API. The models accept text and image inputs and generate text via the Responses API and the Chat Completions API. For service businesses, that matters in two places: drafting and quality control (content, proposals, FAQs), and internal workflows (intake processing, routing tasks, and assembling client ready deliverables). The practical benefit of multimodal inputs is simple: your team can pass the same context you already have, including screenshots and images, then receive text output that can be reviewed and approved. ## Practical Task Split: Sol Vs Luna Trials ### Use Sol For Fast Drafting And Structured Output When your constraint is speed and consistency, trial Sol on work where inputs are already well defined and the output needs to fit a format your team can check quickly. - **First drafts** for service pages, blog posts, and email copy that follow a house style. - **Rewrite passes** that tighten language, remove repetition, or adapt one message for multiple channels. - **Content QA scaffolds**, such as generating an editorial checklist from the draft, generating a short summary for approval, or extracting claims that should be verified before publishing. - **Image to text support** where the team provides an image input (for example a screenshot) and wants a text output that can be turned into a draft, a note, or a checklist. ### Use Luna For Deeper Reasoning And Review Passes When the business risk comes from subtle errors, inconsistent logic, or missed constraints, trial Luna on tasks that look like review and reasoning, especially where the output is used to guide decisions or client communications. - **Second pass review** on drafts created elsewhere, focusing on internal consistency, missing steps, and unclear assumptions. - **Workflow reasoning** for internal processes, such as turning an intake description into a step sequence that a coordinator can follow. - **Approval support** that produces an approver friendly brief: what the output is, what source context it used (as provided in the prompt), and what still requires a human decision. ### A Simple Pattern That Works For Lean Teams A practical setup is a two stage flow: generate with Sol, then review with Luna. It keeps drafting fast while putting the heavier reasoning where it belongs, in the step that decides whether something is ready for a client or the public. ## Rollout Checklist For Safety Review And Approvals This checklist is designed for small teams where “safety” means output integrity, predictable handling of edge cases, and clear responsibility for what goes live. 1. **Pick two trial workflows** only: one content workflow and one ops workflow. Keep the scope tight so you can see failure modes quickly. 2. **Standardize the input bundle** your team will pass in, using the same fields each time (for example: service, audience, tone requirements, and any image inputs you already use in internal review). 3. **Decide the output boundary**: what the model is allowed to produce as final, and what must remain a draft that requires approval. 4. **Define a review role** for each workflow. One named person owns the final decision to publish or send. 5. **Route tool calling implementations** toward the Responses API, since OpenAI is concentrating agent features there. Keep the implementation consistent so performance and behavior are easier to track across Sol and Luna. 6. **Log every trial run** with the final approved output and the approver notes. The goal is to build an internal library of what “good” looks like for your business. 7. **Set a stop rule** for each workflow. If a run produces an output that fails your approval criteria, pause that workflow and adjust the inputs, instructions, or review steps before running it again. ## Where IN-STORE Fits For SMB Teams If you want to trial GPT-6 models without turning your marketing into a science project, the fastest path is to treat content and lead capture as one managed system: guided intake, consistent publishing, local visibility, and workflows that move an inquiry from first contact to booked conversation. IN-STORE supports that kind of AI-assisted execution with human review, so owners and lean marketing leads get momentum without adding internal headcount. GPT-6 Sol and GPT-6 Luna give SMB teams a new set of options inside the API, especially for text and image driven workflows. A tight task split, plus an approvals path, is what keeps experiments productive. If you want help turning trials into a steady AI-assisted content workflow that supports lead generation, explore IN-STORE’s [AI-assisted content marketing and lead generation systems](https://intake.in-store.ca/services). Ready to turn trials into steady AI-assisted content workflows? [Explore IN-STORE services](https://www.in-store.ca/). Source: [OpenAI](https://developers.openai.com/api/docs/changelog)