Claude Team vs. Enterprise: Decide on Controls and Measured Cost, Not the Feature Table

Claude Enterprise used to be bought as "Team, but with pricier flat seats for everyone." The current official pricing works differently: Enterprise is a $20/month seat fee (billed annually), and usage is metered from the very first token at standard API rates. There is no included allowance.

Team, meanwhile, stays seat-based and now mixes two seat types in one organization: Standard ($25/month, or $20 billed annually) and Premium ($125/month, or $100 annually), with roughly five times Standard's included usage.

This changes how the decision should be made. Comparison posts written against the old pricing model simply don't apply to new contracts anymore. Two axes decide it now: (1) do you have requirements only Enterprise can satisfy, and (2) if not, which cost model is cheaper at your organization's actual usage. In that order.

Before any cost math, check whether a feature that only Enterprise has is a hard requirement. If anything below is non-negotiable, the comparison ends here.

• SCIM provisioning — automatic member sync from your identity provider
• Admin-console audit logs — records of user actions, system events, and data access
• The Compliance API — programmatic access to conversations, files, and projects, including Claude Code session content; SIEM and DLP integrations ride on it
• Custom data retention controls
• IP allowlisting, customer-managed encryption keys (CMEK), US-only inference
• A HIPAA-ready configuration (BAA eligibility)

Then check scale. Team caps at 150 seats, so 151 or more means Enterprise whether you like it or not. In the other direction, Enterprise requires at least 20 seats self-serve (a fixed $4,800/year in seat fees alone) and 50 via sales. Below 20 seats, Enterprise isn't on the menu at all.

One correction worth internalizing: SSO and domain capture are [included in Team](https://support.claude.com/en/articles/9266767-what-is-the-team-plan) now. "We need SSO, so Enterprise" — the classic 2025 justification — no longer holds in 2026. If nothing on the list is mandatory and you fit within 2–150 seats, move on to the cost models.

The essential difference between the plans is not features but billing shape. Team is a hybrid: flat seats with an included allowance, metered only past it. Enterprise is pure metering: a seat fee that buys access, and every token billed.

Team's allowance is a combination of 5-hour session windows and weekly caps, applied per member. One person hitting their cap doesn't affect anyone else — but unused headroom can't be shared either. Enterprise has no per-seat plan limits; instead you set spend limits at the organization and user level, and usage stops when they're reached.

Write the per-person monthly direct cost as equations and the structure becomes obvious (annual billing, US list prices throughout).

Enterprise    = $20  + U    (U: that person's metered usage, all products)
Team Standard = $20  + Os   (Os: spend beyond the Standard allowance)
Team Premium  = $100 + Op   (Op: spend beyond the Premium allowance)

For someone who fits inside the Premium allowance (Op = 0), the Enterprise break-even is 20 + U = 100, i.e. U = $80/month. Anyone consuming more than $80/month in metered terms is cheaper on a flat Premium seat; below $80, Enterprise metering wins. The threshold is $80 rather than $100 because Enterprise charges its own $20 seat fee.

Against Standard, there is no positive break-even at all. The annual seat fee is the same $20, and Standard includes an allowance on top (so Os ≤ U always). Under identical conditions, Enterprise structurally cannot undercut Standard on direct cost. Cost alone is essentially never a reason to pick Enterprise. What Enterprise buys is controls, and an operation free of per-seat usage caps.

So what does U = $80 look like in tokens? Using standard per-model API rates, assuming a 5:1 input:output ratio and no caching:

Real Claude Code work stretches these numbers considerably, because prompt caching bills cache reads at roughly 0.1x the normal rate — with 80% of the prompt side served from cache, $80 at Sonnet 5 standard rates is closer to 25M tokens. Model mix and cache hit rate swing the result by multiples, which is exactly why coefficient math on paper isn't good enough. Hence the next section.

You cannot compute U or Os from public documentation — the dollar value of a seat allowance isn't published, and usage varies wildly between organizations. But you can compute them from your own data.

Claude Code exports telemetry (OpenTelemetry) with per-user, per-model token counts — input, output, cache reads, cache creation — and an estimated cost. That estimate is tokens × standard API rates, the same arithmetic Enterprise metering uses. Which means you can put a number on "what would the Claude Code portion cost us on Enterprise metering" from measured data, before changing any contract.

For the Chat and Cowork share, the built-in analytics spend-report CSV has a product column — use that to complement the picture, keeping in mind that on seat-based plans, usage inside the allowance is never expressed in dollars (see the built-in analytics section of the sister post). In engineering organizations the bulk of token consumption sits on the agentic Claude Code side anyway, so measuring that first is the pragmatic move.

1. Measure with telemetry for 4–8 weeks — the first weeks after rollout aren't representative, and a short window will mislead you.
2. Convert to per-user monthly figures; take the median, P90, and max — not the average. Usage concentrates in a few heavy users (the official dashboard literally ships a spend-concentration metric).
3. Count the people above $80/month — that head count is your Premium-seat shortlist. Nobody else needs one.
4. Line up total U against Team seat-mix totals — "n Standard + m Premium + expected overage" versus "$20 × seats + U". From here it's addition.

The same distribution doubles as your seat design if you stay on Team — it tells you exactly who gets Premium. Whichever plan wins, measurement comes first.

No compliance requirements, under 20 seats

Team, full stop. Enterprise starts at 20 seats ($4,800+/year in seat fees alone), so it isn't available to you. Start everyone on Standard and upgrade only the people who hit their caps.

A few dozen seats with a handful of heavy users

Mix seats before anything else. Give Premium to the people measuring above $80/month in metered terms and keep everyone else on Standard plus usage credits. Before jumping to "everyone on Premium" or "the whole company on Enterprise," price out the mix — it's usually the cheapest option by a wide margin.

Weekly caps are stopping actual work

First check whether Premium plus usage credits already solves it: with credits enabled, hitting the allowance no longer blocks anyone — work continues at standard API rates (up to 30% off with bundles). If managing credit balances and cap resets is itself unacceptable to the business, that's when Enterprise's freedom from per-seat limits earns its keep. Design your spend limits as the brake before you migrate.

You're crossing 151 seats

Team's 150-seat cap makes Enterprise mandatory, and the question shifts from "which plan" to "what budget." Each person costs $20 + U, so the distribution of U is the budget. Measure before migrating and both the budget conversation and the initial spend limits come from real data instead of guesses.

You have audit or compliance requirements

Per the checklist: Enterprise. The Compliance API covers Chat conversations, files, and projects plus Claude Code session content, and as of May 2026 it feeds 28 security-product integrations (SIEM, DLP, and more). If you need an official pipe into your existing security operation, this is it. How it relates to telemetry-based audit logs is covered in the FAQ.

You can migrate an existing Team organization to Enterprise in place — conversations, projects, members, and roles all carry over. A few things bite in practice:

• The self-serve migration is irreversible — there is no path back to Team. If you're unsure, measure on Team first and upgrade later.
• There's a window with no access during provisioning, and everyone must log out and back in afterward — schedule it outside working hours.
• Several features come back disabled by default — Skills, code execution, file creation, interactive artifacts, Claude Design, and Claude in Chrome. The data persists, but if re-enabling them isn't on your migration checklist, expect a wave of "it broke after the migration" tickets.
• Unused usage-credit balances roll over to the new usage-based plan.
• Self-serve initially accepts credit card payment only.
• Revisit spend limits immediately after migrating — limits carried over from the Team era can halt newly metered usage far earlier than you intended.

No Enterprise → Team migration is documented. Treat it as a one-way door: easy to walk through, no way back.