Use an eligible launch offer for useful work you already have, then compare the recurring cost. Claude’s existing-subscriber cloud credit has an October 7 claim deadline. OpenAI has now released Pro 500 and dots, and reopened new Pro 200 subscriptions with changed allowance terms. These are published products and conditions, rather than the pre-DevDay rumors this page originally tracked.
October 3, 2026 source check: this updates the September launch-offer record, preserving its original publication date. The September value shortlist retains its other-provider review scope with a narrow OpenAI correction. The Sol 6.1 pricing owner has a separate October 3 economics/access check and retains September Sol 6/Luna benchmark evidence. The cost-saving playbook owns the reusable method; offers do not establish which model produces better work.
What is available, and what remains account-specific
| Item | Evidence and scope | Practical decision |
|---|---|---|
| Claude cloud-session credit | Official announcement: $100 for existing Pro subscribers, $250 for existing Max subscribers; claim by October 7 | Check account eligibility, cloud-only scope and displayed spending expiry |
| Claude limit reset | Official help: refill the offered five-hour or weekly limit; weekly schedule stays unchanged | Inspect the limit and expiry before redeeming |
| Codex banked and incident resets | September grants and incident announcements are historical; official help explains banked resets | Check the current dashboard; do not assume a fresh reset is coming |
| OpenAI Pro 100/200/500 | Published monthly USD tiers; Pro 200 reopened with lower allowance for customers outside its grandfathering window | Check your account’s allowance and renewal terms; Ultrafast is included only in Pro 500 among Pro tiers |
| OpenAI dots | Released, with gradual rollout and plan, age, region and workspace conditions | Confirm access and task accounting before assigning ongoing work |
Claude’s cloud credit is for a particular kind of work
The September 23 ClaudeDevs announcement, rechecked October 3, names existing subscribers: $100 on Pro and $250 on Max. The offer instructions say to claim by October 7, through the claim page or /claim-credit in the CLI, and connect GitHub to start a session. The preceding post limits the grant to one per account and cloud sessions, with usage separate from normal limits. Those posts do not publish a cutoff time or timezone, or a universal spending-expiry date. Check both clocks in your own offer.
This is Claude Code cloud-session credit, rather than cash, a subscription-fee discount or general API credit. The announcement does not establish eligibility for someone subscribing now. Its GitHub step describes the offer instructions; current cloud documentation also supports uploading local repositories.
Anthropic’s September 24 charging clarification says qualifying cloud sessions consume the promotional credit first, then normal plan usage. Ordinary cloud sessions share account limits, parallel sessions consume them proportionately, and the VM has no separate compute fee, according to the cloud documentation. A bounded bug fix or small feature with explicit acceptance checks is a useful trial task; the dollar amount alone is not evidence of completed work.
The promotional-credit terms generally require an active paid plan in good standing, unless an offer overrides that condition. Additional account, product, feature or seat restrictions can apply. Participation authorizes enabling Usage Credit when needed. Cancellation, downgrade to a free plan or suspension can forfeit unused credit unless the offer says otherwise. Do not infer this cloud offer’s spending expiry from the terms’ generic fallback.
Inspect your Usage Credits controls after claiming. Once normal included limits are exhausted, paid overflow uses API rates only when Usage Credits are enabled and funded; monthly caps, alerts and optional auto-reload affect that spending. Enabling the feature does not by itself establish an immediate card charge. Keep paid overflow and renewal cost visible.
A Claude reset refills a limit; it does not add cloud dollars
Anthropic’s limit-reset help, checked October 3, says an eligible reset restores the offered limit: either the five-hour allowance or the weekly allowance. The weekly schedule stays on its usual day and time. Redemption cannot be undone, does not refund billed extra usage, and does not change the credit balance.
Open Settings → Usage on the web or Claude Desktop and inspect the offered limit and displayed expiry before confirming. The redemption button is not currently available in mobile, terminal or IDE clients, although the replenished account allowance applies across those surfaces. Downgrading or cancelling before redemption forfeits the reset. A saved reset and cloud credit can appear together while funding different usage; neither establishes a permanent capacity increase.
A separate artifact usage promotion runs October 1 at 11 a.m. through October 15 at 11:59 p.m. Pacific for Pro, Max and Team. Selected artifact tasks automatically consume 50% less five-hour allowance; weekly usage and Usage Credit charges stay unchanged. It excludes Claude Code, API use and other listed work. This is a temporary usage reduction, not cash, a reset or unlimited artifact work; read the task exclusions before relying on it.
Codex: a banked grant and a separate incident reset
OpenAI’s September 22 launch announcement, rechecked October 3, relays a banked-reset grant for Plus, Pro and Business accounts. The banked-reset help explains redemption and expiry. A full banked reset refills eligible five-hour and weekly allowances and moves the weekly reset date, unlike Claude’s reset. Check the limits applicable to your plan, remaining count and deadline in the dashboard; this does not mean every Pro account has a five-hour limit. The help page’s September 3–4 eligibility examples describe an earlier grant, not the September 22 announcement.
Separately, Tibo Sottiaux’s September 26, 00:07 UTC post announced a reset for paid Codex and ChatGPT Work users following a disruption. At 18:17 UTC the same day, he reported propagation complete. Both posts were rechecked October 3. This remains a provider report about that incident, rather than proof from every account or a standing reset schedule. Plan work against the allowance your account shows.
After DevDay: Pro 500 and dots have published terms
OpenAI’s September 29 DevDay recap confirms the release of dots and the $500 Pro tier with GPT-6 Astra Ultrafast. The September 27 version of this page covered “o” and “Pro Max” reports; current guidance uses the official dots and Pro 500 terms. Official dots documentation identifies GPT-6 Astra; it does not confirm the leaked “Aeon” identity.
The published Pro tiers are Pro 100 at $100, Pro 200 at $200 and Pro 500 at $500 per month, in USD. The current Pro FAQ says new Pro 200 subscriptions are available again with a lower allowance unless grandfathered. Eligibility requires an active Pro 200 subscription at any point from September 22 through September 29, 2026 at 10 a.m. Pacific. Eligible subscribers keep the previous allowance through October 29, 2026 while Pro 200 remains active; after that date they move to the lower allowance at the same $200/month price. Returning on the same account uses that same eligibility window; subscribers outside it receive the updated allowance. The earlier 30-day return rule is superseded. Grandfathering does not add Pro 500 features.
Among Pro tiers, only Pro 500 includes Ultrafast; buying credits on Pro 100 or 200 does not unlock it. Eligible Enterprise/Edu access follows workspace controls and agreements. Work and Codex share usage. OpenAI says Astra Ultrafast generates output tokens up to 8× faster than Astra Standard in Codex; that does not mean 8× faster task completion. It consumes included allowance at 8× Standard for the same model, versus 6× the Standard purchased-credit rate. Compare the work completed within your allowance before adding a $500 recurring bill.
Fast, Ultra and Ultrafast are different controls. Fast is a speed mode: included usage is consumed at 2.5× Standard, purchased credits at 2×. Ultra uses subagents for parallel work rather than setting a token-generation speed. Sol 6.1 supports Standard and Fast at launch; its Ultrafast support is forthcoming. None of these billing multipliers guarantees the same reduction in elapsed task time.
Dots access is gradual: Pro 100/200/500 for users over 18 outside the EEA, UK and Switzerland; Business Premium worldwide; and Enterprise worldwide with administrator enablement. Plus is not listed. Dot conversations do not count toward ChatGPT limits, but Work and Codex tasks count toward those products’ limits as usual. A deeper-work allowance has extended limits for the first month after launch; the page gives no numeric quota or exact expiry. Check your account rather than assuming immediate access or unlimited unattended work.
For a task to try with existing access, use the Dots, Work or Codex chooser: an owner-reported coordination case, a read-only prompt and simpler alternatives. The case does not establish a measured saving or a reason to upgrade for temporary capacity.
AIHackers has not checked signed-in entitlement, checkout or account meters, redeemed an offer, or run a paid comparison for this article. CAR is not-run. Recheck Claude’s claim deadline and each account’s displayed expiry, the artifact promotion’s end, and Pro 200’s October 29 allowance transition. Source archives remain archive-pending.