[BUG] Silent entitlement regression May 22 — 200k cap on Max plan, extra usage charged against plan capacity, 20x token burn rate on identical prompts — no outage, server-side account state, all surfaces
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
<head></head><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Summary</strong></p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">This is a billing and entitlement accuracy issue, not an outage report. All surfaces show green. The service is responding. The problem is that the service being delivered does not match the service being sold. This has been continuous since May 22, 2026 and is unresolved despite status page incidents from that window being marked closed.</p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">This is distinct from #47019, #59872, #54714, and #61692. Those documents report errors and outages. This documents a permanent silent degradation where the system functions correctly — but at a reduced specification that was never disclosed to affected accounts.</p><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>What changed May 22, 2026</strong></p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Prior state (working correctly for months):</p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>Effective context: ~600k (1M window, autocompact at ~60%)</li><li>Extra usage: not being charged</li><li>Token burn per prompt: consistent with historical baseline</li></ul><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Post May 22 state (current, unresolved):</p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>Effective context: 200k hard cap — same limit as six months ago, before Max plan entitlements were extended</li><li>Extra usage: being charged while session plan allocation shows capacity remaining (~40% used)</li><li>Token burn per prompt: orders of magnitude higher than identical prompts before May 22</li><li>Compaction consuming ~300k tokens to compact a 200k window — the compaction call is burning more tokens than the window it is summarizing</li></ul><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">No configuration changes were made by the user. No opt-in changes were made by the user.</p><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Confirmed across all surfaces simultaneously</strong></p>
Surface | Context enforced
-- | --
Claude Code desktop app | 200k
VS Code official extension | 200k
Terminal CLI (/context) | 17.9k / 200k confirmed
<head></head><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>What should happen</strong></p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>Claude Max plan subscribers receive 1M context window as advertised</li><li>Autocompact triggers at ~600k (60% of 1M window), not at 200k</li><li>Extra usage credits are only charged after plan allocation is fully exhausted</li><li>Token consumption per prompt is consistent with historical baseline for identical prompts</li><li>Compaction consumes fewer tokens than the window it is summarizing</li><li><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">hasExtraUsageEnabled</code><span> </span>is only set to<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">true</code><span> </span>through explicit user opt-in action</li><li><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">cachedExtraUsageDisabledReason</code><span> </span>and<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">hasExtraUsageEnabled</code><span> </span>never hold contradictory values simultaneously</li></ul><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>What actually happens</strong></p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>200k hard cap enforced on Max plan despite 1M being advertised</li><li>Autocompact fires within the first prompt of a session — before any file reads or tool outputs</li><li>Extra usage credits charged while session indicator shows plan capacity remaining (~40% used)</li><li>A single prompt that historically consumed ~15k tokens now consumes 300k+</li><li>Compaction process consumes ~300k tokens to compact a 200k window — more than the window it is summarizing</li><li><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">hasExtraUsageEnabled: true</code><span> </span>present in account config without any user action</li><li><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">cachedExtraUsageDisabledReason: "org_level_disabled"</code><span> </span>present simultaneously — server enforces disabled state on every launch, overwriting local state</li><li>Identical behavior confirmed across all three surfaces simultaneously</li></ul><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Error messages / logs</strong></p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">No explicit error messages are surfaced. The service responds normally. Token consumption and context limits are silently wrong. Observable signals:</p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">/context</code><span> </span>output in terminal:<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">17.9k / 200k tokens (9%)</code><span> </span>— 200k cap confirmed</li><li><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">/context</code><span> </span>output in desktop app:<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">21.0k / 200.0k (11%)</code><span> </span>— same cap confirmed</li><li>Extra usage charges appearing in billing while session shows plan capacity remaining</li><li>Session compacting after single prompt with no large file reads or tool outputs</li><li>Account config conflict:<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">"cachedExtraUsageDisabledReason": "org_level_disabled"</code><span> </span>alongside<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">"hasExtraUsageEnabled": true</code></li><li>Anthropic support Fin agent confirmed during live session May 23: incidents between May 18-22 affecting multiple models and services</li><li>Support conversation IDs:<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">215474425771452</code><span> </span>and<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">215474425880773</code></li><li>Extra usage charges exceeded $50 in a single day. A single prompt consumed ~$7. User never opted into overage billing.</li></ul><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Steps to reproduce</strong></p><ol style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>Log into a Claude Max plan account active before May 22, 2026</li><li>Open Claude Code on any surface — desktop app, VS Code extension, or terminal CLI</li><li>Run<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">/context</code><span> </span>— observe 200k cap instead of 1M</li><li>Send a single prompt of moderate complexity equivalent to a standard session priming routine</li><li>Observe autocompact triggering within or immediately after the first prompt</li><li>Check billing — observe extra usage charges despite session plan allocation showing remaining capacity</li><li>Check<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">~/.claude.json</code><span> </span>— observe<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">hasExtraUsageEnabled: true</code><span> </span>alongside<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">cachedExtraUsageDisabledReason: "org_level_disabled"</code></li></ol><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Reproducible on every session start since May 22. No special configuration required.</p><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Is this a regression?</strong></p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Yes. Confirmed regression with precise onset May 22, 2026.</p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Before May 22 (working correctly for months):</p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>Effective context ~600k</li><li>No extra usage charges</li><li>~15k tokens per standard priming prompt</li><li>Compaction triggered at ~60% of 1M window</li></ul><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">After May 22 (every session since):</p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>200k hard cap</li><li>Extra usage charges while plan has remaining capacity</li><li>300k+ tokens for the same prompt — 20x increase</li><li>Compaction triggers within the first prompt</li></ul><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Anthropic status page logged incidents May 20, 21, and 22. All marked resolved. Entitlement state for affected accounts was not restored when incidents were closed. The service is operational. What is not operational is delivering the specification that was sold.</p><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Token burn rate — primary signal</strong></p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">This is not explained by the context cap alone. A 200k window does not cause a 15k prompt to consume 300k tokens. The numbers do not reconcile on their own terms.</p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">What the pattern indicates:</p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>1. The session indicator and the billing counter are reading from different sources.</strong>
Session shows 40% used. Extra usage is simultaneously being charged. Both cannot be accurate unless the billing system is counting tokens the session display is not tracking — most likely the compaction overhead calls.</p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>2. Compaction calls are being billed at a different rate than the session.</strong>
The compaction process makes a separate API call to summarize the conversation. If that call is being routed to a larger model, billed at API rates instead of subscription rates, or if the token counter is double-counting the input sent to the summarizer, the billing meter spikes independently of what the session indicator shows.</p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>3. Something changed in how compaction calls are counted against quota after May 22.</strong>
Before: compaction was transparent to the billing meter or counted within the subscription allocation.
After: compaction calls appear to be counting at full API rate against extra usage credits.</p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">This explains both the billing anomaly and why the session indicator appears normal while the credit balance drains. The 300k token burn on a 15k prompt is not a rounding error. It is a billing routing or calculation change that went into effect May 22 and has not been reversed. This warrants independent investigation separate from the entitlement issue.</p><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Cross-surface confirmation — server-side</strong></p>
Surface | Context enforced
-- | --
Claude Code desktop app | 200k
VS Code official extension | 200k
Terminal CLI | 17.9k / 200k confirmed via /context
<p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Three separate codebases. Three separate update channels. One account. Same broken entitlement state on all three. Public PRs merged to <code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">anthropics/claude-code</code> between May 18-23 contain no changes to billing, context enforcement, or entitlement logic. The regression is in Anthropic's private backend infrastructure. No client update will resolve it.</p><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Claude Code version</strong></p><ul style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; padding-inline-start: 2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><li>Terminal CLI:<span> </span><code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">2.1.116</code></li><li>Desktop app: separate update channel, same account state, same behavior confirmed</li><li>VS Code official extension: separate update channel, same account state, same behavior confirmed</li></ul><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Why this is distinct from existing issues</strong></p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Issues #47019, #34083, #36515 document the 200k cap as an isolated symptom. Issues #59872 and #54714 document billing anomalies in isolation. This report documents the complete picture: context cap, billing order violation, unauthorized entitlement flag change, and token calculation discrepancy appearing simultaneously on the same account on the same date with a confirmed 20x token burn rate increase. The burn rate component and the billing-before-plan-exhaustion component have not been reported together with this level of specificity.</p><hr style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;"><strong>Human escalation denied after being offered</strong></p><p style="font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: pre-wrap; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: initial; text-decoration-style: initial; margin-top: 0.1em; margin-bottom: 0.2em; color: rgb(204, 204, 204); font-family: -apple-system, system-ui, "Segoe UI", Roboto, sans-serif; font-size: 13px; font-variant-ligatures: normal; background-color: rgb(24, 24, 24); text-decoration-color: initial;">Anthropic support (Fin AI agent, Intercom) offered human escalation in response to this issue, then acknowledged it could not fulfill the offer. This occurred seven times across two separate support conversations on May 23. Follow-up emails recommending human contact received no response in five days. Documented in conversation IDs <code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">215474425771452</code> and <code style="font-family: monospace; color: rgb(208, 208, 208); background-color: rgb(60, 60, 60); padding: 2px 4px; border-radius: 3px; word-break: break-word; font-size: 0.9em;">215474425880773</code>.</p><br class="Apple-interchange-newline" style="caret-color: rgb(255, 255, 255); color: rgb(255, 255, 255); font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration-line: none; text-decoration-thickness: auto; text-decoration-style: solid;"><br class="Apple-interchange-newline">
Onset date: May 22, 2026
Duration: 6 days unresolved as of filing
Labels: bug area:cost platform:macos
What Should Happen?
Claude Max plan subscribers should receive 1M context window as advertised
Autocompact should trigger at ~600k (60% of 1M), not at 200k
Extra usage credits should only be charged after plan allocation is fully exhausted
Token consumption per prompt should be consistent with historical baseline for identical prompts
hasExtraUsageEnabled should only be set to true through explicit user opt-in
cachedExtraUsageDisabledReason and hasExtraUsageEnabled should never hold contradictory values simultaneously
Compaction should consume fewer tokens than the context window it is summarizing
What actually happens
200k hard cap enforced despite Max plan subscription
Autocompact fires immediately, often within the first prompt of a session
Extra usage credits charged while session indicator shows plan capacity remaining (~40% used)
A single prompt that historically consumed ~15k tokens now consumes 300k+
Compaction consumes ~300k tokens to compact a 200k window
hasExtraUsageEnabled: true present in account config without user action
cachedExtraUsageDisabledReason: "org_level_disabled" present simultaneously — server enforces disabled state on every launch
Identical behavior confirmed across all three surfaces simultaneously
Error Messages/Logs
Steps to Reproduce
What actually happens
A Claude Max subscription is receiving 200k context instead of the advertised 1M. Autocompact fires within the first prompt before any file reads or tool outputs occur. Extra usage credits are being charged while the session plan indicator shows capacity remaining — approximately 40% used at the time charges were accumulating. A single prompt consumed ~$7 in extra usage credits. Total extra usage charges exceeded $50 in one day. The user never opted into overage billing.
The compaction process itself is consuming approximately 300k tokens to compact a 200k window — more tokens burned by the compaction call than the window it is summarizing. The session display shows normal context usage while the billing meter simultaneously charges extra usage credits. Both cannot be accurate. They are reading from different sources.
Account config contains two contradictory flags set without any user action: hasExtraUsageEnabled: true alongside cachedExtraUsageDisabledReason: "org_level_disabled". The server enforces the disabled state on every launch and overwrites any local state. This account-level inconsistency is present across desktop app, VS Code official extension, and terminal CLI simultaneously — three separate codebases and update channels exhibiting identical behavior on the same account, confirming this is server-side.
Token burn rate — primary signal
Before May 22, a standard priming prompt on this account consumed approximately 15,000 tokens. After May 22, the same prompt, same model, same structure, consumes over 300,000 tokens. That is a 20x increase for identical input. This is not explained by the 200k context cap — a smaller window does not cause a 15k prompt to burn 300k tokens.
The pattern points to compaction calls being billed at full API rate against extra usage credits rather than within the subscription allocation. The compaction process makes a separate API call to summarize the conversation. If that call began routing to a different billing bucket after May 22, the session indicator would show normal usage while the credit balance drains — which is exactly what is being observed. The token counter and the session display are not reading from the same source after May 22. This is a billing calculation change, not a capacity issue, and it warrants investigation independently of the context cap regression.
Claude Model
Not sure / Multiple models
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.116, v2.1.152
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
What actually happens
A Claude Max subscription is receiving 200k context instead of the advertised 1M. Autocompact fires within the first prompt before any file reads or tool outputs occur. Extra usage credits are being charged while the session plan indicator shows capacity remaining — approximately 40% used at the time charges were accumulating. A single prompt consumed ~$7 in extra usage credits. Total extra usage charges exceeded $50 in one day. The user never opted into overage billing.
The compaction process itself is consuming approximately 300k tokens to compact a 200k window — more tokens burned by the compaction call than the window it is summarizing. The session display shows normal context usage while the billing meter simultaneously charges extra usage credits. Both cannot be accurate. They are reading from different sources.
Account config contains two contradictory flags set without any user action: hasExtraUsageEnabled: true alongside cachedExtraUsageDisabledReason: "org_level_disabled". The server enforces the disabled state on every launch and overwrites any local state. This account-level inconsistency is present across desktop app, VS Code official extension, and terminal CLI simultaneously — three separate codebases and update channels exhibiting identical behavior on the same account, confirming this is server-side.
Token burn rate — primary signal
Before May 22, a standard priming prompt on this account consumed approximately 15,000 tokens. After May 22, the same prompt, same model, same structure, consumes over 300,000 tokens. That is a 20x increase for identical input. This is not explained by the 200k context cap — a smaller window does not cause a 15k prompt to burn 300k tokens.
The pattern points to compaction calls being billed at full API rate against extra usage credits rather than within the subscription allocation. The compaction process makes a separate API call to summarize the conversation. If that call began routing to a different billing bucket after May 22, the session indicator would show normal usage while the credit balance drains — which is exactly what is being observed. The token counter and the session display are not reading from the same source after May 22. This is a billing calculation change, not a capacity issue, and it warrants investigation independently of the context cap regression.
This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗