4 min read
OpenAI has just changed the game. Instead of a single "GPT-5.6" model, the company now offers a family of three models built for different needs and budgets: Sol, Terra and Luna. For a Quebec SMB already using Microsoft 365, or exploring artificial intelligence for the first time, understanding this new segmentation is essential before investing time or money in an AI project.
Sol, Terra, Luna: three levels of intelligence, three pricing tiers
According to OpenAI's official announcement, GPT-5.6 is priced per million tokens across three clearly distinct tiers: Sol, the most powerful model, costs $5 USD input / $30 USD output; Terra, positioned as the "balanced" model, costs $2.50 USD input / $15 USD output; and Luna, the most economical option, costs $1 USD input / $6 USD output [1].
Per OpenAI's own help documentation, GPT-5.6 Sol is designed for complex work — coding, knowledge work and research, cybersecurity, science, computer use and design — with an even more capable variant, Sol Pro, reserved for the hardest tasks [5]. Terra, meanwhile, is described as a balanced model for everyday business work: document analysis, automation, and the best quality-to-cost ratio in the lineup, reportedly delivering near GPT-5.5-level performance at roughly half the API cost [6]. Luna rounds out the family as the most affordable option, suited for simple tasks such as content summarization [2].
Early independent benchmarks are telling: Sol reportedly scores 53.6 on the "Agents' Last Exam" evaluation, 13.1 points ahead of Claude Fable 5, while Terra and Luna both outperform Fable 5 as well, at roughly one-sixteenth of the cost [4]. That performance-to-price ratio is exactly what makes this family compelling for organizations that need to justify every technology dollar spent.
Which model should your business choose?
The advice that comes up most often among analysts is simple: start with the least expensive model that meets your need, and only move up if necessary. One SMB-focused guide suggests starting tests with Luna for simple tasks like content summarization, or Terra for more complex analysis, before investing in the flagship Sol model [2]. Another guide echoes the same logic: for most small-business work, Terra or Luna is the right and far cheaper choice, with Sol reserved for genuinely hard tasks [3].
In practice, for a Quebec SMB, that could look like this:
- Luna: summarizing emails, generating template replies, sorting documentation, triaging support requests.
- Terra: analyzing contracts or financial reports, automating repetitive workflows, drafting more elaborate communications.
- Sol: advanced coding, cybersecurity analysis, in-depth research, or tasks requiring complex multi-step reasoning.
Access isn't uniform yet, either. According to Engadget, OpenAI first launched a limited preview of GPT-5.6 to a small group of trusted partners, with a broader rollout expected in the coming weeks [7]. At the same time, OpenAI presents all three models as already available across ChatGPT, Codex, and its API [8]. Either way, caution is warranted: test at small scale before building a mission-critical workflow around a model still in staged rollout.
Beyond the model: the work agent and automation
The GPT-5.6 launch comes with an equally important shift in how AI is used: the ability to delegate multi-step tasks rather than simply asking one question at a time. According to a practical guide on ChatGPT's Work agent, this feature lets you research, draft, send, and follow up across connected tools without babysitting each step [9]. For an SMB, that could mean preparing a quote, drafting a first version of a contract, or compiling a market review, with human oversight only at the key checkpoints.
This is exactly the kind of scenario we're seeing with our own clients: interest in generative AI now goes well beyond a simple chatbot and reaches into automating entire business processes.
What to keep in mind before adopting GPT-5.6 at work
Having three distinct models instead of one is good news for SMBs: it lets you match cost to the actual value of each task, rather than paying premium rates to summarize an email. But this flexibility also adds a layer of decision-making: which model, for which use case, with what data?
That's precisely the kind of strategic — not just technical — decision our clients ask us to help them navigate.
At HiloTech, our HiloIntelligence division deploys private AI solutions directly on our clients' own infrastructure, without sharing data with public models — a particularly important consideration when handling personal or confidential information under Quebec's Law 25. Whether your business is considering integrating a model like Sol, Terra, or Luna into daily operations, or simply wants to understand how these tools fit into your existing Microsoft 365 environment, our senior technicians can assess your actual needs and help you avoid paying for power you don't need.
At HiloTech, no long-term commitment is required: our services run month-to-month, including AI guidance and deployment. Let's talk about your AI project today.
