OpenAI's GPT-5.6 Luna became the default model for Free and Go tier ChatGPT users this week, replacing GPT-5.5 Instant for those tiers. Plus and Pro users get an updated GPT-5.6 Sol with a new thinking-effort slider that lets a person manually choose how much reasoning effort ChatGPT applies to a given response. Unlimited text conversations and a dedicated Think feature are slated to follow. It is the latest step in a run of moves, including a late-July pricing cut and Microsoft defaulting GitHub Copilot to GPT-5.6 Sol for staff, that all point the same direction: OpenAI is optimising for maximum consumer usage.
Two different problems, not two versions of the same race
It is tempting to read OpenAI's free-tier expansion and Anthropic's own cost-visibility guide, published the day before, as competing answers to the same question. They are not. OpenAI is solving for maximising consumer usage: wider free access, a manual effort slider individual users can tune themselves, unlimited conversations. Claude's cost-visibility guide, published a day earlier, is solving for the opposite problem: predictable, governed AI spend for a business running Claude across a team, not maximising an individual's free usage.
Those are genuinely different jobs. A consumer deciding how much of their own time to spend chatting with an AI benefits from a manual effort slider they control themselves. A business trying to keep AI spend predictable across fifty staff, with a finance team that needs to reconcile a monthly invoice, needs something closer to the opposite: admin-set defaults, spend caps that apply automatically, and visibility into who is spending what, not a slider left to each employee's individual judgement.
What AU businesses should actually optimise for
For an Australian business evaluating either platform for team-wide deployment, the free-tier headline is mostly irrelevant. A business is paying for Enterprise or Team-tier access regardless of vendor, and the question that actually determines the total cost of ownership is governance, not the free tier's generosity.
Who sets the model a new conversation defaults to, an admin who has thought about cost-per-task, or each individual staff member choosing on the fly.
Can spend be capped per team or per user before the invoice arrives, or only reviewed after the fact.
Does the platform expose usage analytics broken down by person and team that reconcile against the actual bill, or a single aggregate number.
Is there a documented cost-per-outcome framework, matching model capability to task difficulty, or is everyone defaulting to the most capable model because nobody has set a cheaper default.
Claude's answer to those four questions, entitlements and defaults an admin controls, hard spend caps that apply immediately, usage analytics that reconcile against invoices, and an explicit effort parameter for API-level work, is the governed side of that split. A manual effort slider aimed at individual consumers solves a different problem well; it is simply the wrong tool for a business trying to keep fifty people's combined AI spend predictable.
Where the free-tier race does matter to a business
None of this means free-tier competition is irrelevant to a business owner. It is a reasonable signal of where each lab is investing, and a genuinely widened free tier lowers the bar for staff experimenting with a tool before a business commits to a paid rollout. The mistake is treating a free-tier expansion as evidence about which platform governs cost better once that experimentation turns into a real, team-wide deployment. Those are two separate evaluations, and conflating them is how a promising pilot turns into an ungoverned bill three months later.
There is a second-order effect worth naming too. A workforce that has grown used to a manual effort slider on a free consumer tool will bring that same mental model into a work context if a business does not set its own defaults deliberately. Left alone, staff tend to default to whichever setting feels most capable, not whichever setting the task actually needs, because nobody is paying the bill directly out of their own pocket. That is precisely the behaviour admin-set model defaults exist to correct, and it is a governance gap that shows up faster in a business running dozens of seats than it ever would for a single consumer account.
The Automata AI take
We would treat OpenAI's free-tier push as background noise for the actual decision most AU businesses are making: not which platform is more generous to a free user, but which one gives finance and IT the levers to keep spend predictable once a team of any real size is using it daily. That is the conversation worth having before signing an Enterprise agreement with either vendor, and it is exactly the cost-governance audit we scope for AU SMBs moving past the pilot stage, typically A$3,500 to A$6,000.
Book a brainstorm and we will walk through what a governed Claude deployment would look like for your team size.
