📫 Subscribe | 🗣️ Hire us to speak | 🤝 Sponsor this newsletter | 🎓 Events

Welcome to the Friday edition of our newsletter. We spend Fridays going deeper into tools and trends related to generative AI (and Tuesdays sharing news updates).

ChatGPT now gives users more control over the model, reasoning level, and working mode used for a task. That flexibility is useful, but it also means that simply choosing the most powerful option every time is not necessarily the smartest approach.

A more powerful model may produce a better answer on a difficult problem, but it can also take longer and use more of your plan’s allowance or credits. For routine tasks, a faster model can often produce an equally useful result.

The key is to make two decisions:

ChatGPT now includes three different ways of working.

🗣️ Chat: Choose regular Chat when you need an answer, conversation, brainstorm, explanation, rewrite, summary, or quick analysis. Good examples include:

Regular Chat is still the best starting point for most everyday requests.

💼 ChatGPT Work: Choose Work when you want ChatGPT to complete a longer, multi-step project or produce a finished deliverable.

Work can research and analyze information, use connected apps and files, and create documents, spreadsheets, presentations, reports, analyses, and websites. It can continue through multiple steps, show its progress, ask questions, adjust direction, and request approval before taking important actions.

Work is especially useful when the task sounds more like an assignment than a question. For example:

Work can create or edit native Google Docs, Sheets, and Slides when the appropriate Google Workspace connection is enabled. It can also use reference files and reusable templates to preserve branding, formulas, layouts, tone, and structure.

👩‍💻 Codex: Choose Codex when the main outcome is software development. Codex is designed for writing code, debugging, running tests, reviewing changes, working with repositories, and using developer tools. A request to create a marketing plan belongs in Work. A request to build and test a web application belongs in Codex.

A simple rule:

Need an answer? Use Chat.
Need a finished professional deliverable? Use Work.
Need working software? Use Codex.

Support Innovation Profs by supporting our sponsor:

AI is moving faster than any other technology. New models. New tools. New claims. New noise.

Most people feel like they're behind. But the people that don't, aren't smarter. They're just better informed.

The Future Today is a daily briefing for people who want clarity. In one concise email each day, you'll get the most important AI and tech developments, learn why they matter, and what they signal about what's coming next.

Written for operators, builders, leaders, and anyone who wants to sound sharp when AI comes up in the meeting.

One email. Five Minutes. Stay ahead of 99% of the world.

Join Free Today →

GPT-5.6 is not just one model. It is a family with three capability tiers.

GPT-5.6 Sol: Sol is the flagship and most capable model in the GPT-5.6 family. It is designed for difficult professional work involving research, coding, science, design, cybersecurity, computer use, and complex knowledge work. Choose Sol when:

Examples include developing an AI policy, evaluating competing business strategies, analyzing complicated financial data, conducting a detailed research project, or creating a presentation from several source documents.

Sol is usually unnecessary for a basic rewrite, simple summary, list of ideas, or straightforward factual question.

GPT-5.6 Terra: Terra is the balanced model. It is designed to provide strong intelligence while using fewer resources than Sol. OpenAI describes Terra as competitive with GPT-5.5 while offering a better balance of capability, speed, and cost. For many people, Terra will be the best default model inside ChatGPT Work. Choose Terra when:

Examples include preparing a workshop outline, creating a standard business report, analyzing survey results, drafting a communications plan, or building a presentation from clearly organized source material.

GPT-5.6 Luna: Luna is the fastest and most resource-efficient GPT-5.6 model. It is designed for high-volume and cost-sensitive work while still providing strong general capability. Choose Luna when:

Examples include formatting information, extracting details from documents, creating first drafts, categorizing responses, producing multiple variations, cleaning data, or converting content from one structure to another.

Luna can also be a good first-pass model. You might use Luna to organize information and then use Terra or Sol to analyze the results.

Where These Models Appear: In regular ChatGPT conversations, Terra and Luna are not normally selected directly. GPT-5.6 Sol powers the Medium, High, and Extra High reasoning options on eligible plans.

Sol, Terra, and Luna can be selected inside ChatGPT Work on eligible paid plans. They are also available in Codex and through the OpenAI API, with availability depending on the user’s plan and workspace settings.

Where GPT-5.5 Still Fits: GPT-5.5 Instant remains the default model for fast, everyday ChatGPT conversations. When the model picker says Instant, you are generally using GPT-5.5 Instant. The Medium and higher reasoning settings now use GPT-5.6 Sol on eligible plans.

Do not assume that using GPT-5.5 means accepting a poor result. It remains a capable model and may be the most efficient choice for a large percentage of normal ChatGPT requests.

The intelligence or reasoning control determines how much effort the selected model spends working through the task.

It does not simply make the model knowledgeable or unknowledgeable. It changes how thoroughly the model plans, checks assumptions, follows complex logic, explores alternatives, and verifies its answer. Lower reasoning favors speed and lower resource use. Higher reasoning allows the model to think more completely, but it may take longer and consume more of your available usage.

⬇️ Low: Low uses lighter reasoning and prioritizes speed. Use Low for: Simple formatting, Basic rewriting, Extracting information, Categorizing content, Generating variations, Following a clear template, Routine or repetitive work

A good Low-level task would be: “Turn these 20 event descriptions into a table with columns for date, location, audience, and registration link.”

Low is not ideal when the model must make an important judgment or reconcile conflicting information.

➡️ Medium: Medium balances speed and reasoning depth. It is the best starting point for most professional work. Use Medium for: Writing reports, Developing lesson plans, Comparing options, Analyzing research, Creating presentations, Building project plans, Synthesizing several sources, Handling multiple instructions at once

A good Medium-level task would be: “Review these survey results, identify five important themes, and create an executive summary with recommendations.”

Start with Medium when you are uncertain. Increase the reasoning level only when the task or initial result shows that more depth is needed.

⬆️ High: High gives the model more reasoning depth for complex problems. Use High for: Difficult strategic decisions, Complicated data analysis, High-stakes recommendations, Conflicting or incomplete evidence, Multi-step research, Evaluating risks and tradeoffs, Reviewing legal, ethical, financial, or policy issues, Finding weaknesses in an argument or plan

A good High-level task would be: “Review our proposed AI policy, identify legal, privacy, security, and implementation risks, compare it with the attached policies, and recommend revisions.”

High should not be the automatic choice for every task. It can add time without meaningfully improving straightforward work.

Use GPT-5.5 Instant for fast questions, conversations, rewrites, summaries, brainstorming, and routine help.

Use GPT-5.6 Sol with Medium reasoning for thoughtful analysis, professional writing, planning, comparisons, and most work that benefits from deliberate reasoning.

Use GPT-5.6 Sol with High reasoning for complex, ambiguous, multi-step, or high-stakes problems.

Inside Work, use GPT-5.6 Luna for fast, repetitive, clearly defined tasks.

Inside Work, use GPT-5.6 Terra for most everyday professional deliverables.

Inside Work, use GPT-5.6 Sol when the quality of the analysis or final deliverable matters more than speed or resource use.

The smartest choice is not always the smartest model. The smartest choice is the least expensive, fastest option that can reliably complete the task at the quality level you need.

The best HR advice comes from those in the trenches. That’s what this is: real-world HR insights delivered in a newsletter from Hebba Youssef, a Chief People Officer who’s been there. Practical, real strategies with a dash of humor. Because HR shouldn’t be thankless—and you shouldn’t be alone in it.

Sign Up Free