- ChatGPT Plus top-up: ¥185 ≈ 27.58 USDT (OpenAI list price $20 a month)
- ChatGPT Pro 200 top-up: ¥1,300 ≈ 193.75 USDT; ready-made account: ¥1,349 ≈ 201.05 USDT (list price $200)
- ChatGPT Pro 500 top-up: ¥3,299 ≈ 491.66 USDT; ready-made account: ¥3,399 ≈ 506.56 USDT (list price $500)
- Codex credits: ¥180 ≈ 26.83 USDT per 500 credits on your own Plus or Pro account
- ChatGPT Business: Standard seat ¥220 ≈ 32.79 USDT, Premium seat ≈ 163.94 USDT (min 2 seats, 2 Standard ≈ 65.58 USDT)
- For comparison: Pro 100 top-up 126.68 USDT (list price $100)
What OpenAI shipped, August to October 2026
Everything below comes from OpenAI’s release notes, newsroom and Help Center, read on October 26; press figures are marked as reported. The release notes’ newest entry is dated October 20.
| Date (2026) | What shipped | Why it matters |
|---|---|---|
| Aug 6 | Updated GPT-5.6 Sol in ChatGPT; GPT-5.6 Luna becomes the Free and Go default with unlimited text chats (abuse guardrails apply) | The free tier gets more generous while paid tiers are differentiated by reasoning and agents |
| Aug 10 → 25 | Premium seats for ChatGPT Business: $125 a month ($100 annual), 5x Standard usage, no five-hour limit; generally available August 25 | First per-seat usage tier in a team plan |
| Aug 13 | Ultrafast preview: GPT-5.6 Sol up to 14x faster, up to 750 output tokens a second on Cerebras hardware (API, limited) | Speed becomes something OpenAI sells separately |
| Aug 24 | ChatGPT Business capped at 200 paid seats per subscription; larger teams go to Enterprise | A clear line between self-serve and sales-led |
| Aug 25 | ChatGPT Work can complete tasks on websites you sign in to (Plus and Pro); Jalapeño, OpenAI’s first custom inference chip, posts first results | Agents move from sandbox demos to real accounts; serving costs get a first-party lever |
| Aug 31 | ChatGPT Ads passes a $1 billion annualized run rate in under 200 days | Ads become a pillar that funds the free tier |
| Sep 3 | GPT-6 Astra, first for a limited set of organizations, then Plus, Pro, Business, Enterprise, the API ($10 / $50 per million tokens), Azure and Bedrock | New flagship; inside subscriptions it is metered by allowance |
| Sep 10 | Agents API in public beta: the Codex harness as an API, no fee beyond tokens and tools | The agent runtime itself becomes a product |
| Sep 11–17 | Custom GPTs scheduled for retirement in favour of plugins; automatic Instant→Thinking switching retired on Plus and Pro; ChatGPT for Word on all plans | Fewer legacy surfaces, more explicit model choice |
| Sep 22 | GPT-6 Sol and GPT-6 Luna in ChatGPT Work and Codex; API $2 / $10 and $0.1 / $0.5 | A three-model ladder: Astra, Sol, Luna |
| Sep 29 (DevDay) | Pro 500 ($500, Ultrafast); Pro 200 reopened with a lower allowance for new subscriptions; GPT-6.1 Sol in Work and Codex; Codex Cloud; dots; Space and Pages; plan allowance usable in 16 partner tools | The biggest single repricing and agent launch of the quarter |
| Oct 6–9 | Audio uploads on paid plans; GPT-6 with Intelligent UI in Chat (Sol for paid plans, Luna for Free and Go); Composer predictions beta in Codex for individual Pro users | GPT-6 reaches every plan; Chat answers become interactive |
| Oct 14 | GPT-5.5 retired for ChatGPT sign-in (the API is unaffected) | Older models leave fast, typically within weeks |
| Oct 15–16 | GPT-Live-1.1 voices; Rosalind Academic Plan; ChatGPT Slides for Pro, Business and Enterprise with Space; GPT-6 Pro gets faster responses, better tool use and parallel agents | Pro-and-up features keep stacking |
| Oct 20–22 | Locked chats; OpenAI acquires Orchestra Bio; OpenAI models coming to Google Cloud; ads in 89 countries; new Usage Policies from November 20; Decisions API generally available; Uber rolls ChatGPT Work out to more than 30,000 employees | Distribution and enterprise deals dominate the latest news |
| Oct 30 (next) | Pro 200 grandfathering ends: every Pro 200 moves to the lower allowance at $200 | Same price, less included usage |
The GPT-6 model ladder
In four weeks OpenAI replaced a single flagship with a ladder. The names matter because each one sits in a different place in ChatGPT and costs a different amount of your allowance:
| Model | OpenAI’s positioning | Where in ChatGPT | API price per 1M tokens (in / out) | Codex credits per 1M output |
|---|---|---|---|---|
| GPT-6 Astra (Sep 3) | Most capable model for complex work | Chat as GPT-6 Pro on Pro $100, Pro $200, Business and Enterprise; Work and Codex on all paid plans | $10 / $50 | 1,250 |
| GPT-6.1 Sol (Sep 29) | Near-Astra performance at a lower cost | ChatGPT Work and Codex only, not Chat | $2 / $10 (cached $0.10) | 250 |
| GPT-6 Sol (Sep 22) | The everyday GPT-6 model (our summary) | Instant to Extra High in Chat on paid plans; Work and Codex | $2 / $10 | 250 |
| GPT-6 Luna (Sep 22) | Most efficient model for focused, high-volume tasks | Chat on Free and Go; Work and Codex | $0.1 / $0.5 | 12.5 |
Two details trip people up. GPT-6.1 Sol, the model Uber made its ChatGPT Work default, is still not selectable in Chat. And OpenAI’s October 7 announcement puts Free and Go on GPT-6 Luna, while the Help Center page on models still said GPT-5.6 Luna when we read it today; trust the model menu in your own account. Benchmarks in OpenAI’s launch posts (for example GPT-6.1 Sol matching Astra on DeepSWE v1.1 at about a fifth of the cost) are the vendor’s own tests.
Retirements are fast: GPT-4.5 left ChatGPT on June 26, OpenAI o3 was scheduled out for August 26, gpt-5.4 left ChatGPT-signed-in Codex on August 31, GPT-5.3-Codex-Spark on September 14 and GPT-5.5 on October 14. If you script a model name, expect to change it a few times a year.
Plans: three Pro tiers and the October 30 change
| Plan | List price | What it adds (official) | Usage vs Plus (reported, not official) |
|---|---|---|---|
| Free | $0 | Unlimited GPT-6 text chats (abuse guardrails), limited Codex and Work; may show ads | — |
| Go | $8 | More messages, uploads and images; may include ads | — |
| Plus | $20 | GPT-6 thinking at Medium and High; GPT-6 Astra and GPT-6.1 Sol in Work and Codex with five-hour windows; credits for overflow | 1x |
| Pro 100 | $100 | GPT-6 Pro, Extra High, deep research, a dot; no five-hour limit in Work and Codex | ~5x |
| Pro 200 | $200 | More than Pro 100; new allowance for every subscription after October 29 | ~10x (was ~20x) |
| Pro 500 | $500 | Highest Pro allowance; Ultrafast for GPT-6 Astra and GPT-6.1 Sol in ChatGPT Work and Codex | ~25x |
| Business Standard | $25 / seat ($20 annual) | Workspace, admin controls, no training on business data by default; 15 Pro messages a month | Same five-hour estimates as Plus |
| Business Premium | $125 / seat ($100 annual) | 5x Standard, no five-hour limit, 50 Pro messages a week, dots | ~5x (our derivation) |
Grandfathered Pro 200 subscriptions keep the old allowance through October 29; from October 30 every Pro 200 runs on the lower new allowance at the same $200. OpenAI’s own explanation is that the new allowance reflects “increasingly efficient models”, and the Help Center no longer prints multiples or GPT-6 Pro message counts (it did until late September). The reported numbers come from Engadget, a statement by OpenAI’s Thibault Sottiaux relayed by kingy.ai, and Unite.AI. A one-off grant of 62,500 credits to affected subscribers appears only in second-hand reports. Pro is monthly only: no annual plan, no paying months ahead.
Agents: ChatGPT Work, Codex and dots
- One meter for agent work. ChatGPT Work and Codex share the same usage, credits and limits. Fast mode uses included usage at 2.5x and Ultrafast at 8x (credits at 2x and 6x). OpenAI calls these consumption rates, not speed guarantees.
- Codex Cloud (September 29) runs tasks in reusable cloud environments from desktop, web or phone, and keeps going while your computer sleeps. Pro, Business and Enterprise get 4 vCPU / 16 GiB machines.
- Dots are always-on agents on GPT-6 Astra with their own cloud computer and access to more than 4,000 apps through plugins. For the first month dots do not count toward Pro, Business and Enterprise allowances; on our reading that runs to about October 29, and OpenAI has not yet published the terms that follow.
- GPT-6 Pro in Chat gained parallel agents on October 16: several agents work on parts of a task and you can open each one’s work.
- For developers, the Agents API (September 10) sells the Codex harness itself with no fee beyond tokens and tools, and the Decisions API (October 22) answers bounded questions on GPT-6 Luna at $0.10 per million input tokens with no output charge.
The usage data points the same way. In OpenAI’s small-business report, agentic output tokens from ChatGPT Work and Codex made up two thirds of small-business output tokens in August, up from one third in April. Uber took ChatGPT Work from a 1,000-person pilot to more than 30,000 employees.
Our read: five trends for the next 6–12 months
These are our judgments, not OpenAI’s plans. Each lists the signals behind it, how confident we are, and what would prove it wrong.
1. Prices stay put; allowances and speed move
Our read: OpenAI will keep round list prices and change what they include, selling speed and extra capacity as multipliers and credits.
- New Pro 200 subscriptions got a lower allowance at the same $200, explained by more efficient models.
- Pro 500 sells Ultrafast as an 8x usage burn; Fast is 2.5x on every paid plan.
- The Help Center stopped publishing message counts and multiples, and OpenAI now runs its own inference chip next to a multi-vendor compute portfolio, which gives it room to reprice capacity.
Confidence: likely. Wrong if: OpenAI publishes fixed weekly caps again or raises a headline price such as Plus.
2. Agents become the product; Chat is the entry point
Our read: the features that separate plans will increasingly be agent capacity (dots, cloud tasks, parallel agents) rather than chat models.
- Dots, Codex Cloud, Work and the GPT-6 Pro parallel-agent update all landed in five weeks, and dots are Pro and Business Premium only.
- Agentic tokens doubled their share of small-business output between April and August.
- OpenAI sells the same harness to developers through the Agents API.
Confidence: likely. Wrong if: the post-launch dots terms make them a costly add-on few people keep, or OpenAI folds dots back into ordinary chat.
3. Enterprise and distribution get the most attention
Our read: more of OpenAI’s launches will target companies and come through other platforms.
- Business Premium seats and the 200-seat cap that hands bigger teams to Enterprise.
- OpenAI models in Google Cloud’s Model Garden in addition to Azure and Bedrock; Uber’s 30,000-seat Work rollout; Slides, Space and Pages limited to Pro, Business and Enterprise.
- New Usage Policies from November 20 make model routers and AI gateways seek OpenAI’s approval, which tightens control of indirect access.
Confidence: likely. Wrong if: the next big launches are consumer-first and available on Plus and Free at once.
4. A two-sided consumer business: ads below, premium agents above
Our read: the free and $8 tiers will be funded more by ads, while paid tiers stay ad-free and gain agent features.
- Ads passed a $1 billion run rate in under 200 days and now run in 89 countries, only on Free and Go.
- Free got unlimited GPT-6 text chats; Plus, Pro and Enterprise are promised to stay ad-free.
- Reuters reported on October 9, citing a source, that OpenAI put its September annualized revenue at almost $50 billion, up from $20 billion at the start of the year (reported, not confirmed by OpenAI).
Confidence: likely for ads on Free and Go; speculative for any ads above Go. Wrong if: ads appear on Plus.
5. Faster model turnover, with Chat lagging Work and Codex
Our read: new models will reach Work and Codex first, Chat later, and older models will keep leaving within weeks.
- GPT-6.1 Sol is still Work and Codex only; GPT-6 reached Chat two weeks after Work and Codex.
- Five retirements between June and October, the last (GPT-5.5) one month after GPT-6 Sol arrived.
Confidence: possible for GPT-6.1 Sol reaching Chat within six months; likely for continued retirements. Wrong if: OpenAI promises long support windows for named models in ChatGPT.
What Pro buyers should expect
- Your allowance can change at the same price. Read Settings → Usage, not old multiples, and judge a plan after a week of real work.
- Buy month by month. Pro is monthly only, so match the tier to the month: Pro 500 for a crunch, Pro 200 or Pro 100 otherwise.
- Use credits for spikes. They cover Codex, ChatGPT Work and the Office add-ins, last 12 months and do not add GPT-6 Pro messages or unlock Ultrafast.
- Watch the dots terms. The free first month is temporary; dots tasks in Codex and Work already count.
- Keep Ultrafast for moments that matter. At 8x it drains a Pro 500 allowance quickly.
What to buy now
| If you… | Buy | Price today |
|---|---|---|
| Mostly chat, research and occasional Codex | ChatGPT Plus | ¥185 ≈ 27.58 USDT |
| Use Codex or ChatGPT Work daily, want GPT-6 Pro and a dot | Pro 200 (top-up or ready-made) | ¥1,300 ≈ 193.75 USDT / ¥1,349 ≈ 201.05 USDT |
| Hit the new Pro 200 limit early every week, or need Ultrafast | Pro 500 | ¥3,299 ≈ 491.66 USDT / ¥3,399 ≈ 506.56 USDT |
| Run out only in some weeks | Codex credits on your Plus or Pro | ¥180 ≈ 26.83 USDT per 500 |
| Buy for a team (two or more people) | ChatGPT Business seats (Premium = 5 Standard) | ¥220 ≈ 32.79 USDT per Standard seat |
How it works here: top-ups go on your own account, and the sign-in data is checked before you pay, so a failed check costs nothing. Ready-made accounts are hand-delivered in about 30 minutes to 4 hours after you verify a delivery email; change the password on first sign-in. Every order covers 30 days and is paid in USDT (TRC20 or BEP20). Ready-made accounts are not refundable once delivered; if anything goes wrong, contact the live chat or Telegram promptly.
Risk, briefly: OpenAI’s terms apply per user, and access from unsupported regions may be restricted. ClaudeMax is independent, not OpenAI or an authorized reseller; OpenAI sets the allowances and may change them again.
For the other two labs in this series, read Anthropic and Claude in October 2026 and the cross-brand AI industry trends for Q4 2026.