Update — July 16, 2026: Third Anthropic reset this month — @ClaudeDevs refilled 5h + weekly limits; Codex and Cursor matched the same day. Sequel: Claude, Codex, Cursor limit reset day.
Update — July 13, 2026 (afternoon): Tibo denied Sol thinking-budget nerfs and added ~10% usage + banked resets scaling to 7M users. New: Rob Hallam hospitalized after Fable limit stress · ChatGPT Work vs Codex guide · Tachikawa Fable string theory anecdote. 5-hour cap still removed · Fable extended to July 19.
On July 9, 2026, two frontier labs launched within hours — and both reset rate limits within 48 hours. Anthropic refilled Fable 5 5-hour and weekly buckets via @ClaudeDevs. OpenAI hit ~2× traffic on GPT-5.6 Sol and @thsottiaux reset Codex + ChatGPT Work limits twice in one day.
These are not unrelated goodwill gestures. They are retention insurance in the most competitive model week since the July 1 Fable restore — the cliff was July 12, then moved to July 19 on July 12 evening.
TL;DR — why both reset
| Company | Reset | Stated trigger | Real driver |
|---|---|---|---|
| Anthropic | 5-hour + weekly (Jul 9) | "Enjoy more Fable!" — @trq212 | July 19 cliff (extended Jul 12) · GPT-5.6 + Grok pressure |
| OpenAI | Codex + ChatGPT Work ×2 (Jul 10) · 5h cap removed + weekly reset (Jul 12) | "demand surge" — @thsottiaux | GPT-5.6 GA · Sol Ultra · 5h gone Jul 12 |
explainx.ai read: Limit resets are cheap churn prevention. Permanent cap raises are expensive. One-time refills let each lab survive launch week without repricing subscriptions.
The competitive backdrop (July 9–12)
Three models dropped in the same news cycle:
| Model | Ship | Headline |
|---|---|---|
| GPT-5.6 Sol / Terra / Luna | Jul 9 | GA in ChatGPT, Codex, API — full comparison |
| GPT-5.6 Sol Ultra | Jul 10 | Cycle Double Cover Conjecture proof — 64 subagents, <1 hour |
| Grok 4.5 | Jul 8–9 | SpaceXAI + Cursor coding push |
X aggregators summarized the horse race: Fable 5 ~60 pts vs Sol ~59 on composite frontier scores — Sol wins coding agents and runs at roughly ⅓ the cost per task. That is the pressure Anthropic felt when @ClaudeDevs posted:
We've reset 5-hour and weekly rate limits for all users.
Thariq (@trq212) amplified: "Enjoy more Fable!"
Why Anthropic reset Fable limits
1. The July 12 credit cliff
Anthropic's July 8 email extended included Fable through Sunday July 12, 11:59pm PT. Monday July 13 → all Fable usage on usage credits ($10/$50 per M tokens).
A limit reset July 9 gives developers a full refreshed bucket for the final promo weekend — maximum Fable exposure before billing changes.
2. OpenAI forced Anthropic to keep bundling Fable
@theo (22h after launch):
I would like to thank OpenAI for putting out a model exactly good enough to force Anthropic to keep bundling Fable in the Claude Code plan
Theo's point: GPT-5.6 Sol is competitive enough that Anthropic cannot drop Fable from subscriptions without mass downgrade risk — but not so dominant that Anthropic surrenders on pricing yet.
3. "Third place at $200/month"
Sir, if we pull Fable out of the subscription, our next best model is Opus 4.8 and even Grok beats that now, let alone GPT Sol. we'd be charging $200/month for third place.
That X post captures Anthropic's retention math:
Without Fable in-plan:
├── Opus 4.8 = best included model
├── Grok 4.5 > Opus on coding (per community benchmarks)
├── GPT-5.6 Sol > Opus on agents (per OpenAI July 9 thread)
└── $200/mo Max → hard to justify
Resetting limits + extending promo = delay the churn moment while advisor patterns teach cheaper Fable usage.
4. Capacity, not charity
Fable limits exist because inference is scarce. A reset means Anthropic allocated extra burst capacity for launch week — same playbook as the July 1 restore reset. It does not mean permanent unlimited Fable.
Why OpenAI reset ChatGPT / Codex limits twice
1. GPT-5.6 launch surge
@thsottiaux, July 10:
Hello beautiful people! We have reset usage limits across Codex and ChatGPT Work. And another one will come later in the day. Rejoice.
Context from the same thread: traffic ~2× previous peak after July 9 GA. Sol Ultra, merged desktop + ChatGPT Work, and upgraded Codex features drove unprecedented burn.
2. Pro subscribers hit walls fast
@theo, ~1h before this post:
gpt-5.6-sol without hitting limits — I've burned over $200,000 of tokens with gpt-5.6-sol. It's a great model. That said, it's a bit too easy to hit your limits on the $200 Codex Pro sub. OpenAI has been generous about resets…
The pattern: best model in years → worst limit UX → public complaints → emergency reset. Twice in one day signals OpenAI's ops team was still catching up to demand at noon PT.
3. The Fable cost comparison
our team's first week of having both sol and fable — the crazy thing is 30% of our cost was on fable
@sama replied skeptically — "30% of the cost was on fable at these levels of usage?" — but the thread matters for why OpenAI resets: teams benchmarking both stacks burn double quota. OpenAI clears Sol limits so the comparison does not end on "I ran out of GPT before I could finish the eval."
4. Launch-week optics
Same week as Apple sues OpenAI, Musk–Altman feud, and bio bounty. Limit rage on X would drown Ethan Knight's math proof. Resets buy narrative space.
Side-by-side — reset mechanics
| Anthropic (Fable) | OpenAI (GPT-5.6) | |
|---|---|---|
| What reset | 5-hour + weekly plan buckets | Codex + ChatGPT Work limits |
| Who | All Claude users (per @ClaudeDevs) | Paid / Work tiers (per Tibo thread) |
| Frequency | Once (Jul 9) + prior Jul 1 restore | Twice (Jul 10) |
| What did NOT change | July 13 credits cliff | Sol/Terra/Luna pricing |
| Competitive target | Keep Fable in subscription vs Sol | Keep builders on Codex vs Claude Code |
What developers should do this weekend (July 12)
If you're on Claude Max/Pro:
- Burn refreshed Fable quota before Sunday 11:59pm PT — last included window
- Test Fable advisor + Sonnet executor to stretch quota
- Plan July 13+ for credits-only Fable or Opus/Sonnet default
If you're on Codex Pro / ChatGPT Work:
- Sol for agents · Terra/Luna for cost — tier guide
- Expect another reset only if traffic stays at 2× — not guaranteed
- Compare total $/task vs Fable before July 13 billing change
If you're evaluating both:
Run the same repo on Fable (this weekend) and Sol (post-reset) — @thdxr's 30% cost split is the metric enterprises will cite in Q3 procurement.
What happens after the resets stop
| Date | Anthropic | OpenAI |
|---|---|---|
| Jul 12 | Last day included Fable at 50% weekly cap | GPT-5.6 rollout continues tier-by-tier |
| Jul 13 | Credits-only Fable | Normal limit policy — no more launch-week resets promised |
| Ongoing | Will Fable return to subscriptions? | GPT-5.6 international waves |
Neither company has announced permanent unlimited access. Jul 9–12 resets were launch-week exceptions; OpenAI went further July 12 by removing the 5-hour cap (temporarily).
Related on explainx.ai
- ChatGPT drops 5-hour limit — Tibo weekly reset Jul 12
- Will Fable extend again? Polymarket ~25% theory
- Fable extended to July 19 — latest @claudeai
- Fable 5 goes usage-based today — July 12 deadline (superseded)
- GPT-5.6 vs Fable 5 — full benchmark comparison
- Fable extended through July 12 — credits July 13
- When will GPT-5.6 be available — rollout waves
- Grok 4.5 launch — third-place pressure on Opus
- Fable 5 rate limits and credits guide
- Codex vs Claude Code harness comparison
- Apple sues OpenAI — same news week
Sources: @ClaudeDevs · @trq212 · @thsottiaux · @sama · @theo · @eknight
Limit policies change by account tier and region. Verify your plan in Claude and OpenAI account settings — social posts are announcements, not contracts.