Chat and Work Are One Tab Now: A Selection Guide for ChatGPT's Shared Quota Pool
On November 2, 2026, OpenAI confirmed that ChatGPT's long-separated Chat and Work tabs will merge into a single surface — and, more consequentially, that ChatGPT and Codex will draw from the same usage allocation pool. The tab merge is a UI story. The pooled quota is a budgeting, procurement, and workflow-design story, and it is the part that should drive your decision.
This guide treats the announcement as a purchasing and configuration problem: what the options are, which criteria actually discriminate between them, and what to pick for six common scenarios. Where OpenAI has not published specifics (exact pool sizes, per-model burn rates, overage handling), those items are flagged as verify before committing rather than guessed at.

Figure: Editorial context image from the source page (preview only; rights remain with the publisher) · Credit / further reading: techcrunch.com
What actually changed
Three things matter for planning:
- One tab, one context. Conversation history, projects, and connectors are no longer split by "is this chat or is this work." Continuity improves; so does the risk of a personal thread inheriting an enterprise connector.
- One pool. Tokens, agent runs, and Codex task time come out of a single allocation instead of two. An afternoon of Codex refactoring now visibly reduces the budget for document drafting.
- One admin surface. Seat management, retention policy, and spend controls converge — which is good for governance and bad for teams that relied on keeping two cost centers apart.
The practical consequence: you can no longer buy your way out of a coding-heavy month by pointing at a separate Work budget. Plan selection becomes an exercise in forecasting total mixed usage, not two independent forecasts.
The one number that now decides everything
Under a shared pool, your effective capacity is:
(Total allocation) ÷ (blended burn rate of your actual mix)
A team that is 80% chat and 20% Codex has a very different blended burn rate from a team that is 40/60 — even at identical seat counts and identical plan tiers. Before the merge, those two teams could be sized independently. Now they cannot.
This is why the selection criteria below weight mix tolerance and overage behavior above raw headline allowance. A generous pool with no soft landing is worse than a modest pool with predictable overflow.
The options
Option A — Individual tiers (Free / Plus / Pro)
Single-seat plans, self-serve, no admin console worth the name. Post-merge, these tiers are where the shared pool is most visible to the user, because there is no organizational buffer: when the pool empties, you stop working. Pro-tier seats are the only individual option with enough headroom to run Codex meaningfully alongside daily chat.
Choose when: you are one person, your usage is bursty but bounded, and you value zero procurement friction.
Option B — ChatGPT Business (self-serve team)
The middle path. Shared workspace, seat-based administration, SSO, and — critically for the merged world — some form of pooled or centralized billing so one heavy Codex user doesn't strand five colleagues. Data from business seats is excluded from consumer training by default.
Choose when: you are 5–200 people, you have a mixed chat/code population, and you want one invoice without a sales cycle.
Option C — Enterprise / Education with custom controls
Custom allocation, negotiated overage, domain verification, audit logs, residency options, and the governance machinery regulated buyers need. The shared pool becomes a contract term rather than a UI meter.
Choose when: compliance, retention, or data residency requirements exceed what self-serve terms provide.
Option D — API / Platform plus Codex CLI
You buy tokens directly, meter them yourself, and route chat and code through the same key. There is no "pool" to exhaust in the UI sense — there is a bill. Maximum control, maximum engineering overhead, and no bundled workspace features.
Choose when: you already have an internal AI gateway, or your volume makes per-seat economics indefensible.
Option E — Hybrid: seats for people, API for pipelines
Seats for humans doing interactive work; API keys for anything automated (CI jobs, batch summarization, scheduled agents). This keeps autonomous workloads from silently draining human-facing allocation — the single most common failure mode after a pool merge.
Choose when: you have any automation at all. Which is to say, most teams above ten people.
Criteria table
Ratings are relative within this comparison, not absolute. Cost and capacity figures are not published per-tier in a form this guide can responsibly quote — treat the cost row as directional and confirm current numbers before signing.
| Criterion | A: Individual | B: Business | C: Enterprise/Edu | D: API/Platform | E: Hybrid |
|---|---|---|---|---|---|
| Pool scope | Per seat, hard stop | Seat-based, limited central pooling | Contractual, negotiable | Metered spend, no cap | Split: seats + metered |
| Mix tolerance (chat ↔ Codex) | ●●○○○ | ●●●○○ | ●●●●● | ●●●●● | ●●●●○ |
| Admin & policy controls | ○○○○○ | ●●●○○ | ●●●●● | ●●○○○ (build it) | ●●●●○ |
| Data-use / retention guarantees | ●○○○○ | ●●●●○ | ●●●●● | ●●●●○ | ●●●●○ |
| Overage behavior | Hard stop or upsell prompt — verify | Soft limit / admin top-up — verify | Pre-negotiated | Pay-as-you-go | Bounded by API budget |
| Setup effort | Minutes | Hours | Weeks | Days–weeks | Days |
| Cost predictability | High | Medium | High | Low | Medium |
| Best fit | Individuals | 5–200 person teams | Regulated / large | Engineering-led orgs | Teams with automation |
How to read it: the table's center of gravity is mix tolerance versus predictability. Options D and E tolerate any chat/code ratio but hand you a variable bill. Options B and C give you a fixed bill but require you to forecast the ratio in advance. There is no row where you get both, and any vendor claim otherwise deserves a pilot.

Figure: HEXOPS original illustration inspired by the topic (not a copy of the source article imagery) · Credit / further reading: techcrunch.com
Recommendation by scenario
1. Solo knowledge worker (writing, research, occasional code). Option A. The merge barely changes your life; you were already using one pool of your own attention. The one adjustment: check whether Codex access on your tier is gated, since a single long agent run can now consume a visible chunk of a shared monthly allowance.
2. Indie developer or very small studio. Option A Pro, escalating to E as soon as you script anything. The failure mode here is a nightly CI job quietly eating the allocation you needed for morning interactive work. Separate the automated path onto a key early.
3. Startup, 10–60 people, mixed roles. Option B, with an explicitly designated admin who watches the pool weekly for the first two months. Because the old Chat/Work split is gone, expect the first month's data to be unrepresentative — you have no historical baseline for a merged pool. Budget for one top-up cycle and treat it as the cost of measurement.
4. Regulated enterprise (finance, health, public sector). Option C, and negotiate overage before rollout. The question to put to the vendor is not "how big is the pool" but "what happens at 100%, in writing, and who gets notified." A merged pool inside a compliance perimeter without a documented exhaustion path is an audit finding waiting to happen.
5. Agency or consultancy with spiky client work. Option E. Client work is inherently lumpy and you cannot re-forecast seats monthly. Keep humans on seats for the interactive 70%, route batch deliverables and any agentic pipelines through metered access so a single deadline week doesn't strand the rest of the roster.
6. Education and research groups. Option C where available; Option B otherwise, with per-course or per-lab sub-workspaces if the admin layer supports it. Mixed usage is the norm here, not the exception, so pool visibility for a lab lead matters more than raw headroom.
A three-week migration checklist
- Instrument before you commit. Run one week on your current setup and log chat versus Codex consumption separately. You need a ratio, not a vibe.
- Set a soft alarm at 70%. Whatever the admin console offers, agree internally on who acts when the pool crosses 70% and what they do.
- Move automation off seats. Anything scheduled, batch, or agentic goes to a key. This alone prevents most post-merge surprises.
- Re-baseline at week three. The first merged month will over- or under-state steady-state usage. Do not renegotiate seats on week-one data.
- Write down the exhaustion policy. Even for a five-person team. "We ran out and nobody knew who to ask" is the predictable cost of a unified pool.
The bottom line
The tab merge is cosmetic; the pool merge is the decision. Choose on mix tolerance versus cost predictability, not on headline allowance — because in a shared pool, allowance without a stated overage path is a number, not a guarantee. For most teams between ten and two hundred people, that points to Option B with Option E's discipline layered on top: seats for humans, keys for machines, and one named person watching the meter.
Sources
- TechCrunch — "OpenAI is finally merging ChatGPT's Chat and Work tabs": https://techcrunch.com/2026/11/02/openai-is-finally-merging-chatgpts-chat-and-work-tabs/
- OpenAI product and plan documentation — confirm current tier names, allocation sizes, overage handling, and data-use terms before purchase (verify; figures were not published in a quotable form at the time of writing).
- Vendor admin-console documentation for the specific tier under evaluation — specifically the pooling, alerting, and top-up behaviors referenced in the criteria table.



