All three sell a 5x tier at $100 and a $200 tier, but ChatGPT Pro $200 drops to 10x on 30 Oct. We compare coding agents, limits and bundles by persona.
The short answer: if most of your day happens inside a coding agent, buy the plan that owns the agent you already use. Claude Max for Claude Code, ChatGPT Pro for Codex, Google AI Ultra for Antigravity. All three sell a $100 tier with roughly 5x their base plan's usage and a $200 tier, so the price no longer separates them. The agent, the flagship model and the extras do. One multiplier is moving: ChatGPT Pro $200 drops from 20x to 10x Plus on October 30, 2026, while Google's and Anthropic's $200 tiers stay at 20x.
Google made this a three-way race at I/O on May 19, 2026. It added a $99.99 AI Ultra tier and cut the top Ultra plan from $249.99 to $199.99. A lot of comparison pages still list Ultra at $250. That price is gone.
This guide was last checked against Google's Gemini Apps help center, the Google One AI plans page, the Antigravity blog and pricing page, the Google Developers Blog, Claude's pricing page and help center, OpenAI's Codex pricing docs and TechCrunch's reporting on the Pro pause on September 24, 2026, updated on September 30, 2026 for the ChatGPT Pro changes OpenAI announced on September 29, and updated again on October 14, 2026 for Google's October 9 Gemini app changes.
Updated October 14, 2026: Google changed the Gemini app on October 9, and the changes touch this guide. Per Google's Gemini Apps usage limits page, users without a Google AI plan now default to Auto, where most prompts go to Flash-Lite, and with smart model selection off every reply uses Flash-Lite. The multipliers on that page are standard limits with no plan, 2x for AI Plus, 4x for AI Pro and "5x or 20x higher than AI Pro" for AI Ultra, refreshing every five hours until a weekly limit. Each model also gets low, medium and high effort levels, and higher effort uses more of your limit. Two more changes come only from 9to5Google's report, and do not appear legibly on a Google page: AI Plus ($4.99) loses Gemini Pro (per 9to5Google) and keeps Flash-Lite and Flash, with each subscriber's date sent by email, and AI Pro ($19.99) gets Deep Think. Treat the Deep Think change as unconfirmed: 9to5Google reported Deep Think coming to AI Pro, but Google's help page has since been reported to list it for AI Ultra only. Google gives no absolute prompt counts or weekly caps, so the Ultra-versus-Pro comparison below is unchanged. The ChatGPT Pro $200 cut from 20x to 10x Plus on October 30 is as described above and in ChatGPT Pro $100 vs $200 explained.
Quick summary
- Same multipliers, until October 30. Google AI Ultra, Claude Max and ChatGPT Pro all come in a ~5x tier at $100 and a ~20x tier at $200, measured against each vendor's $20 plan, so the $200 tier costs twice as much for four times the usage. From October 30, ChatGPT Pro $200 is 10x Plus, which makes OpenAI's $200 step twice the usage for twice the price.
- Claude Max is the pick for Claude Code users. Opus 5.5 runs inside the plan limits, and Max is the only Claude plan where Fable 5.1 is included (up to 50% of your weekly limit).
- ChatGPT Pro is the pick for Codex users. It has the clearest published limits: GPT-6 Astra in Codex is estimated at 25–225 local messages per five hours on Pro $100 and 100–900 on Pro $200. Pro $200 reopened to new buyers on September 29 after a pause that began September 10, but its allowance falls from 20x to 10x Plus on October 30, 2026. OpenAI also added a $500 Pro tier with Ultrafast Codex, which neither rival matches. Which tier fits Codex work is covered in ChatGPT Pro $100 vs $200 for Codex.
- Google AI Ultra is the pick if you live in Google's tools or want the bundle. Even the $99.99 tier comes with 20 TB of storage, YouTube Premium, $40 a month in Google Cloud credits and 10,000 Flow credits. Deep Think is Ultra-only (reported to be expanding to AI Pro; unconfirmed).
- Gemini CLI is gone for consumer plans. It stopped serving free, Pro and Ultra users on June 18, 2026. The terminal agent on Google plans is now Antigravity CLI (
agy).
The master table: the three $100 tiers
| Google AI Ultra ($99.99) | Claude Max 5x ($100) | ChatGPT Pro ($100) | |
|---|---|---|---|
| Usage vs base plan | 5x AI Pro | 5x Claude Pro per session | 5x Plus |
| Coding agent | Antigravity 2.0 app + Antigravity CLI (agy) | Claude Code | Codex (CLI, IDE, cloud) |
| Agent limit model | Shared token quota across Gemini models, drawn down at API prices; non-Gemini models have a separate fixed limit | 5-hour sessions plus weekly caps | Per-model message ranges per 5 hours, weekly limits may apply |
| Flagship coding models | Gemini 3.8 Flash, Gemini 3.1 Pro | Opus 5.5 (Fable 5.1 up to 50% of weekly limit) | GPT-6 Astra, GPT-6 Sol, GPT-5.6 family |
| Deep-reasoning mode | Deep Think (Ultra-only) | Fable 5.1 | Pro reasoning |
| Published agent numbers | Relative only (5x Pro) | Relative only | Astra: 25–225 local messages per 5 h |
| Bundle | 20 TB storage, YouTube Premium, $40/mo Cloud credits, 10,000 Flow credits | None | None |
| Availability | Open | Open | Open |
The master table: the three $200 tiers
| Google AI Ultra ($199.99) | Claude Max 20x ($200) | ChatGPT Pro ($200) | |
|---|---|---|---|
| Usage vs base plan | 20x AI Pro | 20x Claude Pro per session | 20x Plus until Oct 29, 10x from Oct 30 |
| Coding agent | Antigravity + agy | Claude Code | Codex |
| Flagship coding models | Same as $99.99 tier | Same as Max 5x | Same as Pro $100 |
| Published agent numbers | Relative only (20x Pro) | Relative only | Astra: 100–900 local messages per 5 h (20x table) |
| Bundle | 20 TB+ storage, YouTube Premium, $40/mo Cloud credits, 25,000 Flow credits | None | None |
| Availability | Open | Open | Open again since Sept 29, 2026 |
Two notes on these tables. First, the $200 tiers mostly buy more room. The models, agents and features stay the same as at $100. Second, Google's own pages describe Ultra storage as "20 TB and above" without saying what the $200 tier gets. Several third-party trackers say 30 TB. I have not been able to confirm that on a Google page.
All three vendors price the $200 tier at 2x the $100 tier for 4x the usage, until ChatGPT Pro $200 moves to 10x Plus on October 30.
What the multipliers actually mean
All three vendors describe limits as multiples of a smaller plan. They measure them in different ways, and those differences matter more than the headline numbers.
Google: compute-based, 5-hour refresh, weekly ceiling
The Gemini Apps help center and usage limits page says AI Plus gets 2x the standard limits (what users without a plan get), AI Pro gets 4x, and AI Ultra gets "5x or 20x higher than AI Pro limits depending on your subscription." Your allowance refreshes every five hours until you reach a weekly limit. Google does not publish absolute prompt counts. AI Pro and Ultra both get a 1 million token context window in the Gemini app.
Antigravity works a little differently. After the May 19 plan changes, Gemini Flash and Gemini Pro share one token quota that is "drawn down as per API pricing," so a Pro-model call uses more of it than a Flash call. The $100 Ultra tier gets 5x the Pro plan's token quota and the $200 tier gets 20x. Non-Gemini models have their own fixed rate limits, which Google attributes to capacity constraints. The Antigravity pricing page lists Claude Sonnet and Opus 4.6 and gpt-oss-120b alongside Gemini 3.8 Flash and Gemini 3.1 Pro. AI credits no longer come with the base plans and now work only as paid overage.
In short, Ultra buys a lot of Gemini capacity. It does not buy a lot of Claude capacity inside Antigravity. If you want Claude models, pay Anthropic directly.
Anthropic: per-session multiples plus weekly caps
Claude's pricing page lists Max "from $100/month," with 5x or 20x more usage than Pro in each five-hour session and weekly caps on top. When Opus 5.5 launched on September 22, Anthropic's developer account said Claude Code's 5-hour session limits were going up 20% and that Opus 5.5 would go about 25% further within those limits because it is cheaper to run. Anthropic's Fable help article says Max plans include Fable 5 and Fable 5.1 for up to 50% of weekly limits. On Pro, Fable runs only on paid usage credits.
For the full breakdown, see Claude Max $100 vs $200 for Claude Code and our new Claude Pro vs Max after Opus 5.5 guide.
OpenAI: the only vendor with per-model numbers
OpenAI's Codex pricing docs publish estimated local-message ranges per five hours, and local messages and cloud chats share one allowance:
| Model in Codex | Plus ($20) | Pro $100 | Pro $200 |
|---|---|---|---|
| GPT-6 Astra | 5–45 | 25–225 | 100–900 |
| GPT-6 Sol | 15–150 | 70–700 | 300–3,000 |
| GPT-6 Luna | 350–3,000 | 1,750–14,000 | 7,000–56,000 |
| GPT-5.6 Sol | 10–100 | 50–500 | 200–2,000 |
The Pro $200 column is the 20x table, valid until October 29; OpenAI had not published 10x ranges when this was updated. Weekly limits "may also apply." The ranges are wide because a long agentic turn over a big repo costs much more than a quick edit. More detail is in ChatGPT Pro $100 vs $200 explained and our new Codex and GPT-6 limits by ChatGPT plan.
The ChatGPT Pro $200 pause, and what replaced it
On September 10, 2026, OpenAI paused new sign-ups and upgrades to Pro $200 (Pro 20x), including upgrades from Pro $100. TechCrunch reported that demand for GPT-6 Astra, launched September 3, was straining capacity.
On September 29 OpenAI reopened the plan and cut it. From October 30, Pro $200's included usage in Codex and ChatGPT Work falls from 20x to 10x Plus, and GPT-6 Pro chat messages fall from 200 to 100 a week. Existing subscribers keep 20x until October 29. It also launched Pro $500, with the highest allowance and Ultrafast, up to 8x faster GPT-6 Astra token generation in Codex; see ChatGPT Pro 500: What $500 a Month Buys.
This changes the $200 comparison. From October 30, if you want 20x of your vendor's base plan at $200, only Claude Max 20x and Google AI Ultra at $199.99 offer it; ChatGPT Pro $200 gives 10x. The multipliers use different baselines, so that is not the same as saying ChatGPT gives half the usage of the others. For Codex-specific sizing, see ChatGPT Pro 5x vs 20x for Codex coding.
Deep reasoning: Deep Think vs Fable vs Pro reasoning
Each vendor keeps a slower, more expensive mode for its top tiers:
- Google Deep Think is Ultra-only, on both the $100 and $200 tiers (reported to be expanding to AI Pro; unconfirmed). Google describes it as "maximum parallel reasoning" and says it can take a few minutes to respond. It lives in the Gemini app, not the coding agent. Gemini 3.5 Pro, announced at I/O, was still unreleased as of late-August reporting. A newer Deep Think probably depends on it, so today's Ultra gets you Deep Think on current models.
- Claude Fable 5.1 is included on both Max tiers up to half your weekly limit and uses that limit faster than other Claude models. Opus 5.5, according to Anthropic's launch messaging, performs at Fable 5.1's level for less. Many Claude Code users will leave it as the default and save Fable for hard problems.
- ChatGPT Pro reasoning and GPT-6 Astra are included on every Pro tier.
For a head-to-head on model quality, see GPT-6 Astra vs Claude Fable 5.1 vs Opus 5.5 for coding.
Per-dollar value
On paper, Google and Anthropic price the step the same way. $100 buys 5x and $200 buys 20x, so the $200 tier costs half as much per unit of usage. If you run out of the $100 tier by midweek, the $200 tier is better value than buying extra usage. ChatGPT Pro worked the same way until October 29. From October 30, its $200 tier buys 10x, the same price per unit as the $100 tier, so upgrading buys a higher ceiling rather than a better rate.
The multipliers are measured against different baselines, though, so "5x" doesn't mean the same amount across vendors:
- Google's 5x is 5x AI Pro, which is already 4x the free tier.
- Anthropic's 5x is 5x Claude Pro per session, with weekly caps on top.
- OpenAI's 5x is 5x Plus, and it is the only one you can check in messages.
The bundle is where Google stands apart. List price on YouTube Premium Individual plus $40 a month in Cloud credits is a real chunk of the $99.99, before you count 20 TB of storage. If you already pay for YouTube Premium and a big Google One storage plan, the effective price of Ultra's AI is well under $100. If you'd never use any of it, the bundle doesn't matter.
Google AI Ultra vs Pro: is the jump worth it?
Google AI Pro costs $19.99 a month. For that you get 4x the standard Gemini limits, 5 TB of storage, YouTube Premium Lite, $10 a month in Cloud credits, 1,000 Flow credits and "higher" Antigravity rate limits. The $99.99 Ultra tier adds 5x Pro's limits, Deep Think, full YouTube Premium, 20 TB of storage, $40 in Cloud credits and 10,000 Flow credits.
Upgrade when Antigravity or agy keeps hitting the weekly ceiling, or when you need Deep Think. Before May, the only step up from Pro was $249.99. Now it's $99.99, so most people should try that tier before the $199.99 one.
Google AI Pro vs Claude Pro
At about $20 each, these two are closer than the premium tiers. Claude Pro ($20 a month, or $17 a month billed annually) includes Claude Code with Opus 5.5 and Sonnet, but not Fable 5.1 except through usage credits. Google AI Pro includes Antigravity and agy with Gemini 3.8 Flash and 3.1 Pro, a 1M-token context window in the Gemini app, and the storage and YouTube extras. If terminal coding is the main job, pick by agent: Claude Code or Antigravity. If you want the most for $20 overall, Google AI Pro has the bigger bundle.
Which plan should you pick?
You already use Claude Code every day
Claude Max 5x ($100), then Max 20x if you hit weekly caps. Opus 5.5 raised the effective session allowance, so check whether 5x is enough before upgrading. Antigravity lists Claude models, but on a separate, tighter quota, so Google Ultra doesn't replace a Max plan for this use. If you are choosing between agents, compare Claude Code vs OpenAI Codex limits.
You already use Codex
ChatGPT Pro $100. At 25–225 Astra messages per five hours it covers most solo developers. Pro $200 is open again, but from October 30 it gives 10x Plus, the same price per unit of usage as $100, so move up only when weekly limits actually bind. If you already have Pro $200, re-measure after October 30. Pro $500 is for developers who want Ultrafast speed.
You work mostly in Google's tools, or want Antigravity
Google AI Ultra $99.99. You get Antigravity 2.0 and agy with 5x Pro's Gemini quota, Deep Think, and a bundle that effectively pays for part of the plan. Move to $199.99 only if the weekly ceiling actually gets in your way. Gemini CLI users on consumer plans had to move to agy in June whatever they chose. See our coding-agent CLIs comparison for how it stacks up.
You want more than one frontier model
Split $100 + $100. Claude Max 5x plus ChatGPT Pro $100 costs the same as one $200 plan and covers both main terminal agents. Add Google AI Pro at $19.99 if you want Antigravity and Gemini on the side. We made the full case in ChatGPT Pro vs Claude Max at $200. A split works well when one agent reviews the other's work, and badly when you'd rather have one very deep allowance.
For the wider field, including Cursor Ultra, see Cursor Ultra vs Claude Max 20x vs ChatGPT Pro and the $6–$300 decision matrix for solo developers.
Caveats
- None of these limits are fixed. All three vendors change them without much notice, and weekly caps are applied at the vendor's discretion.
- Google and Anthropic publish multiples, not counts. Only OpenAI gives per-model message estimates, and even those are wide ranges.
- Jules limits aren't published. Google says Pro gets "expanded" and Ultra gets "higher" task and concurrency limits, with no numbers.
- Google changed the Gemini app on October 9. Free users are on Flash-Lite through Auto, and AI Plus reportedly loses Pro. If you compare the $20 tiers, remember that AI Pro is now the cheapest Google plan that keeps Gemini Pro.
- Gemini's next Pro model is late. If Gemini 3.5 Pro ships, Ultra's value moves with it. Our Gemini 3.7 Flash coverage covers the delay.
- Prices are US list prices before tax. Regional pricing differs a lot. India, for example, has its own rupee tiers.
FAQ
Google AI Ultra vs Claude Max vs ChatGPT Pro: which is best for developers?
It depends on your agent. Claude Max for Claude Code, ChatGPT Pro for Codex, Google AI Ultra for Antigravity. All three have a ~5x tier at $100 and a $200 tier. Google's and Anthropic's $200 tiers are 20x; ChatGPT Pro $200 is 20x until October 29 and 10x from October 30. It reopened to new sign-ups on September 29 after a three-week pause.
How much does Google AI Ultra cost now?
$99.99 a month for the 5x tier and $199.99 a month for the 20x tier in the US. Google cut the top tier from $249.99 at I/O on May 19, 2026. Pages quoting $250 are out of date.
Google AI Ultra vs Claude Max 20x: which gives more coding usage for $200?
Neither vendor publishes absolute numbers, so there is no exact answer. Ultra $199.99 gives 20x AI Pro's Gemini token quota in Antigravity. Claude Max 20x gives 20x Claude Pro per session in Claude Code, with weekly caps. Choose based on which models you want to run.
What are Antigravity's pricing and limits?
Antigravity has a free individual tier with basic weekly limits. On paid plans, AI Pro ($19.99) is the baseline, Ultra $99.99 gets 5x Pro's token quota and Ultra $199.99 gets 20x. Gemini models share one quota drawn down at API prices. Non-Gemini models such as Claude have a separate fixed limit.
Is Deep Think included in the $100 Google AI Ultra plan?
Yes. Google says Deep Think is exclusive to AI Ultra and applies the same Ultra feature set to both tiers. Only the usage multiplier and bundle size differ. (9to5Google reported it coming to AI Pro; unconfirmed.)
Does Gemini CLI still work with Google AI Pro or Ultra?
No. Gemini CLI stopped serving free, AI Pro and AI Ultra users on June 18, 2026. Google replaced it with Antigravity CLI (agy). Enterprise Standard and Enterprise licenses and paid API keys are unaffected.