Use this when you're deciding how to pay for Claude Code for a team, or want to know what overage costs before you enable it. Prices and limits were checked against Anthropic's pricing, docs and support pages on October 6, 2026; treat them as a point-in-time snapshot.
Claude Code has two billing models underneath it: seat-based subscriptions with usage windows, and per-token API billing with no ceiling. This October 2026 edition covers two months of changes (there was no September edition).
Two dates that changed your bill
- September 14, 2026: the weekly-limit promotion ended. The +50% weekly boost that ran from May (and was extended past the August 19 date we reported last month) expired on September 13. From September 14, standard weekly limits in Claude Code are permanently 25% higher than the pre-promotion baseline on Pro, Max, Team and seat-based Enterprise. On our arithmetic, going from 150% to 125% of baseline is roughly 17% less weekly headroom than in August, so heavy users will hit weekly limits sooner and reach for usage credits more often.
- September 22, 2026: Claude Opus 5.5 became the default. Opus 5.5 launched at $4 input / $20 output per 1M tokens, 20% below Opus 5, with cache reads at $0.20 (0.05x input). Claude Code's docs now list it as the default model on every plan and on the Anthropic API. Anthropic also raised five-hour usage limits on Pro, Max, Team and seat-based Enterprise with the launch (no figure published) and gave subscribers a one-off rate limit reset they can save and use later.
Quick answer
- Lightest option: Claude Pro at $20/mo ($17/mo annual) includes Claude Code with the smallest allowance.
- Daily driver: Max 5x at $100/mo and Max 20x at $200/mo, priced per individual.
- Teams of 2–150: Team plan with standard seats ($25/mo, $20 annual) and premium seats ($125/mo, $100 annual, 5x standard usage), mixable per person.
- Large orgs: Enterprise at $20/seat/month (billed annually) plus all usage at API rates.
- Automation and CI: the API, per token. Anthropic's published enterprise average is ~$13 per developer per active day and $150–$250 per developer per month, with 90% of users under $30 per active day.
- Overage: usage credits at standard API rates, off by default, with monthly spend limits. Prepaid bundles cut the cost by 10–30%.
The ways to pay for Claude Code in October 2026
| Option | Price | What you get | Best for |
|---|---|---|---|
| Pro | $20/mo ($17 annual) | Base allowance shared across Claude and Claude Code; Fable models only via usage credits | Occasional sessions |
| Max 5x | $100/mo | 5x Pro usage per 5-hour session; Fable models up to 50% of weekly limits | Full-time individual use |
| Max 20x | $200/mo | 20x Pro usage per 5-hour session; Fable models up to 50% of weekly limits | Long agentic sessions, parallel work |
| Team standard seat | $25/mo ($20 annual) | More usage than Pro, admin controls; Fable via usage credits | Most of a team |
| Team premium seat | $125/mo ($100 annual) | 5x standard-seat usage; Fable up to 50% of weekly limits | Your heaviest engineers |
| Enterprise | $20/seat/mo (annual) + usage | All consumption at API rates; SCIM, audit logs, Compliance API, Analytics API | Orgs wanting usage-shaped billing |
| API (Console) | Per token | No usage windows, no ceiling; auto-created "Claude Code" workspace for tracking and limits | Automation, CI agents, bursty use |
Pro, Max and Team are allowance plans where overage is a separate, opt-in charge; Enterprise and the API are metered from the first token. The Team plan now starts at 2 seats rather than 5, so very small teams can use premium seats instead of individual Max plans.
What changed since August
| Date (2026) | Change | Effect on spend |
|---|---|---|
| Early September | Claude Fable 5.1 released at $10/$50, cache reads $0.25 | Same plan treatment as Fable 5; cheaper cache reads on long sessions |
| September 1 | Sonnet 5's $2/$10 introductory price made standard; the planned move to $3/$15 was cancelled | The 50% rise we flagged in August did not happen |
| September 14 | Weekly limits permanently +25% vs pre-promotion; +50% promotion ended Sept 13 | ~17% less weekly headroom than August |
| September 22 | Claude Opus 5.5 at $4/$20; becomes Claude Code default; five-hour limits raised | Lower API cost on the default model |
| September 28 | Claude Sonnet 5.5 at $2/$10; the sonnet alias now resolves to it on the Anthropic API | Cheapest strong option for subagents and routine work |
Per-token prices fell while weekly subscription headroom shrank: API teams got cheaper, subscription teams got more overage-prone. Correction to our August edition: the Sonnet 5 rise to $3/$15 we listed for September 1 was cancelled.
How the usage limits work
Subscription plans meter Claude Code with two overlapping caps. Both count usage from every Claude surface: chat, Desktop, Cowork and Claude Code share one pool.
- A rolling 5-hour session window, sized by plan tier and raised again on September 22.
- A weekly limit that resets at a fixed time each week, now 25% above the pre-May baseline.
On Max plans and on premium seats, Fable 5 and Fable 5.1 can use up to 50% of the weekly limit at no extra cost. They draw from the same pool as other models and use it faster. On Pro and Team standard seats, Fable models run entirely on usage credits at $10/$50 per 1M tokens. Fable is never the default in Claude Code; you have to select it with /model fable, and interactive sessions show a consent prompt where credits apply.
Switching models with /model won't restore access after a session or weekly limit. Anthropic doesn't publish hour or token equivalents for any tier, so any "X hours per week" figure is an estimate.
What happens when you hit a limit, and what overage costs
Claude Code stops and shows when the window resets. You can wait, upgrade, or spend usage credits:
- Priced at standard API rates, billed separately from the subscription, with a $2,000 daily redemption limit.
- Off by default, with a monthly spend limit you set.
- Bundles discount it: $50 for $45 (10% off), $250 for $200 (20% off), $1,000 for $700 (30% off). Bundles are available on Pro, Max and Team. Pro/Max can buy up to $2,000/month in bundle value and Team owners up to $3,000. Bundle credits apply only after plan limits are exhausted.
- New expiry rule: from September 10, 2026, usage credits expire six months after purchase in certain jurisdictions, including Japan.
/usage-creditsin the terminal opens billing settings, or sends a request to your admins on Team/Enterprise.
The cache detail that makes overage cost more than the rate card suggests: prompt-cache lifetime is one hour on a subscription but drops to five minutes once you're drawing on usage credits, unless you set the TTL yourself. More cache misses mean more context reprocessed at full input rates.
On Team and Enterprise, admins set credit spend limits at organization, group and member level. A per-person cap is the most effective control: an agent left running over a weekend becomes a blocked request instead of a budget overrun.
Paying with the API instead
On an API key there are no usage windows, just tokens. Current list prices for the models Claude Code uses:
| Model | Input ($/1M) | Output ($/1M) | Cache read ($/1M) | 5-min cache write ($/1M) |
|---|---|---|---|---|
| Claude Fable 5.1 | $10.00 | $50.00 | $0.25 | $12.50 |
| Claude Opus 5.5 (default) | $4.00 | $20.00 | $0.20 | $5.00 |
| Claude Opus 5 | $5.00 | $25.00 | $0.50 | $6.25 |
| Claude Sonnet 5.5 / Sonnet 5 | $2.00 | $10.00 | $0.20 | $2.50 |
| Claude Haiku 4.5 | $1.00 | $5.00 | $0.10 | $1.25 |
Opus 5.5 and Sonnet 5.5 now cost the same per cached token, so in long sessions the gap is mostly output. Thinking bills as output and can't be turned off on Opus 5.5, Sonnet 5.5 or the Fable models; lower it with /effort (both 5.5 models default to medium). Fast mode on Opus 5.5 costs $8/$40, US-only inference adds 1.1x, and agent teams use roughly 7x the tokens of a standard session. Set a spend limit on the auto-created "Claude Code" Console workspace before you need it.
Worked example: a 10-engineer team, seats vs API
Illustrative example. Token volumes and active days below are StackSpend assumptions chosen to sit inside Anthropic's published ranges. They are not measured data. Prices are October 6, 2026 list rates, monthly billing.
The team: 2 heavy agentic users (long sessions, subagents, 20 active days/month), 5 moderate users (daily pairing, 18 days), 3 light users (a few sessions a week, 8 days). All on Opus 5.5 at default effort.
| Profile | Assumed daily tokens (cache read / uncached input / cache write / output) | API cost per active day | Per month |
|---|---|---|---|
| Heavy | 40M / 2M / 1.5M / 0.6M | $8.00 + $8.00 + $7.50 + $12.00 = $35.50 | × 20 = $710 |
| Moderate | 12M / 0.8M / 0.5M / 0.2M | $2.40 + $3.20 + $2.50 + $4.00 = $12.10 | × 18 = $217.80 |
| Light | 4M / 0.3M / 0.2M / 0.08M | $0.80 + $1.20 + $1.00 + $1.60 = $4.60 | × 8 = $36.80 |
The moderate profile sits near Anthropic's $13 average; the heavy profile sits above the $30 line 90% of users stay under. Pure API: 2 × $710 + 5 × $217.80 + 3 × $36.80 = $2,619.40/month.
| Setup | Fixed cost/month | Assumed overage | Total |
|---|---|---|---|
| A. All API (Console) | $0 | n/a | $2,619 |
| B. Team: 2 premium + 8 standard | 2 × $125 + 8 × $25 = $450 | $380 (see below) | $830 |
| C. Team: 10 premium | 10 × $125 = $1,250 | ~$0 | $1,250 |
| D. Enterprise | 10 × $20 = $200 | All usage at API rates: $2,619 | $2,819 |
Setup B's overage is the key assumption: $150/month of credits for each heavy user (exceeding the premium weekly limit about one week a month) and $40 each for two moderate users. Anthropic doesn't publish seat allowances in tokens, so check this against your own /usage data. On annual billing, B's seats cost $360.
For interactive use, seats win: setup B stays cheaper than pure API until monthly overage passes about $2,170 ($2,619 − $450). Setup C buys predictability for $420 more than B. Enterprise tracks the API curve plus seat fees, so you choose it for governance, not unit cost. Light users are the marginal case: $36.80/month of API usage against a $25 seat that also covers chat. If your heavy users mostly run unattended agents or CI, that work belongs on the API regardless.
Which option should you pick?
- One person coding most days: Max 5x. Move to Max 20x only after
/usageshows you hitting 5x limits most weeks, which is more likely since September 14. - A team: Team plan, standard seats by default and premium seats for the top 20–30% of users. Enable usage credits with per-member spend limits, then review seat tiers monthly. A standard-seat user whose credits regularly exceed $100 should move to a premium seat, since the $100 price difference already covers that overage.
- Agents, CI and scheduled jobs: the API, with a spend limit and a rate limit on the Claude Code workspace.
- Model choice: keep Opus 5.5 as default, put subagents on Sonnet 5.5 or Haiku, and make Fable an explicit choice.
See also per-seat or per-token AI coding tool pricing and AI coding tool budget per engineer in 2026.
Keeping Claude Code spend visible
Anthropic's tooling covers each surface:
/usageshows plan bars, credit spend against your limit, a 24-hour/7-day breakdown by skill, subagent, MCP server and loop, and prompt-cache hit rate.- Team/Enterprise admin: a daily per-user, per-model spend report with CSV export (once credits are on); Enterprise adds the Analytics API.
- Console (API): workspace spend and rate limits, plus per-member Claude Code reporting.
- OpenTelemetry:
CLAUDE_CODE_ENABLE_TELEMETRY=1exports per-user cost metrics on every setup, including Bedrock and Foundry.
Each surface reports only itself. In setup B, finance sees a Team invoice, a separate usage-credit charge and perhaps a CI API workspace, with Cursor or Copilot seats on another bill.
StackSpend puts these on one timeline: Claude Code and Anthropic API spend per developer and per model, alongside Cursor and other AI coding tools, with same-day anomaly alerts when someone's daily spend jumps. That way a looping agent or a post-September-14 overage run gets caught after one day, not at month end. See Claude Code cost analysis, Anthropic cost monitoring, Claude cost monitoring and AI coding tool cost monitoring. If the API side is already high, start with why your Claude API bill is so high.
FAQ
How much does Claude Code cost in October 2026?
Pro $20/month, Max 5x $100, Max 20x $200, Team $25 (standard) or $125 (premium) per seat, or Enterprise at $20/seat/month billed annually plus usage at API rates. On the API, Anthropic's published average is about $13 per developer per active day and $150–$250 per month.
Did Claude Code weekly limits go down in September 2026?
Compared with August, yes. The +50% promotion ended September 13, and from September 14 weekly limits are permanently 25% above the pre-promotion baseline on Pro, Max, Team and seat-based Enterprise. That is higher than before May, but about 17% below what users had during the promotion. Five-hour limits were raised separately on September 22.
What model does Claude Code use by default now?
Claude Opus 5.5, on every subscription plan and on the Anthropic API, at medium effort. It costs $4/$20 per 1M tokens on the API, 20% less than Opus 5. The sonnet alias resolves to Sonnet 5.5 ($2/$10) on the Anthropic API. Fable 5.1 is never the default and has to be selected explicitly.
How is Claude Code overage priced?
Through usage credits at standard API rates, billed separately from the plan, with a $2,000 daily redemption limit. Prepaid bundles discount them: $50 for $45, $250 for $200, $1,000 for $700. Prompt-cache lifetime drops from one hour to five minutes while you're on credits, so overage reprocesses more context than the same work did inside the plan.
Is a Team plan or the API cheaper for a 10-person team?
For engineers coding interactively, the Team plan almost always is. In our illustrative example, 2 premium and 8 standard seats cost $450/month before overage, against about $2,619/month of equivalent API usage, so overage would need to pass about $2,170/month before the API is cheaper. Unattended agents and CI should still go on the API.
Can admins cap Claude Code overage per person?
Yes. On Team and Enterprise, admins set usage-credit spend limits at organization, group and individual-member level, and requests stop when one is reached.
Related reading
- Claude Code Cost Options and Overage (August 2026), the previous edition
- The Best Coding Models in October 2026
- Per-Seat or Per-Token: Which AI Coding Tool Pricing Model Suits Your Team
- AI Coding Tool Budget per Engineer (2026)
- Why Your Claude (Anthropic) API Bill Is So High
- AI Coding Cost Overruns: The Failure Modes
- LLM API Pricing Index: per-1M-token rates, updated daily
References
- Claude Pricing: Pro, Max, Team, Enterprise and API
- Claude API Docs: Pricing (model table, Sonnet 5 footnote, fast mode, data residency)
- Anthropic: Introducing Claude Opus 5.5 (September 22, 2026)
- TechCrunch: Anthropic releases Sonnet 5.5 (September 28, 2026)
- BleepingComputer: Anthropic's September 14 weekly-limit change, quoting Anthropic
- Claude Code Docs: Manage Costs
- Claude Code Docs: Model Configuration (defaults, aliases, effort)
- Claude Support: Claude Fable Models on Your Plan
- Claude Support: Usage Credits for Paid Claude Plans
- Claude Support: Buy Usage Bundles
- Claude Support: What Is the Max Plan
- Claude Support: What Is the Team Plan
- Claude Support: How Usage and Length Limits Work
- Claude Support: View Usage Analytics for Team and Enterprise Plans
- Claude Code Docs: Monitoring Usage (OpenTelemetry)
