On this page
November 20, 2026 · What changed
Qwen Cloud. The October $50 quote was for Alibaba Cloud’s international Coding Plan. The November price change applies to a separate product, the Qwen Cloud Individual Token Plan. Check the product name on your bill before comparing prices. Qwen Cloud now lists four Individual Token Plan tiers. On November 9 it raised all four monthly allowances and reduced its Pro price. Lite is $6 for 20,000 credits, Essential $10 for 45,000, Standard $18 for 80,000 and Pro $43 for 200,000. These are listed monthly prices before new-user offers. When both plan and extra-pack credits run out, service pauses. If you already subscribe, check the top-up for your current cycle before buying a pack. Qwen plan details
GitHub Copilot. Pro is $10/month and Pro+ is $39/month. If you work in a cloud sandbox, check your Actions budget too. GitHub's November 19 update sets November 23 as the start of sandbox billing against the included Actions compute entitlement. Usage above it follows your billing settings. This matters even if your Copilot subscription price stays the same. Copilot plans · GitHub's billing announcement
Which bill should you compare?
For an IDE subscription, Cursor lists Plus at $8/month and Premium at $30/month. Third-party model usage is purchased separately. For terminal work, check the Codex allowance included with your ChatGPT plan or Claude Code access on Claude Pro. Claude Pro's US monthly reference price is $20. API keys have their own billing. Cursor prices · Codex plans · Claude Pro pricing · Claude Code access
Buying from Japan
ChatGPT's web checkout supports JPY and credit or debit cards. Check your local price at checkout; Apple and Google manage their own subscriptions. API credits are purchased separately in USD. For Claude Pro, compare the actual Japanese checkout total: Anthropic specifies 10% Japanese consumption tax for its services to Japanese customers. OpenAI billing in Japan · Anthropic's Japanese tax notice
The November review covers the services above. The full October comparison below retains the prices, availability and dates recorded at the time.
October 2026 edition · historical prices and limits
If you want an AI coding agent without spending much, ChatGPT Go at $8 a month already runs Codex. Chinese plans go even lower: StepFun's Step Plan starts at $6.99. Most people settle around $20, where ChatGPT Plus, Claude Pro and Google AI Pro cost almost the same. What sets them apart is whose models you get and how your usage is counted. For heavy daily use, expect to pay $100 or more. A few recent changes are worth knowing about; you'll find them at the end.
All prices are monthly, in US dollars and before tax, unless we say otherwise. We checked every one against the vendor's own pricing page on October 8, 2026, and you'll find the sources at the bottom.
Comparison at a glance
OpenAI Codex (comes with ChatGPT plans)
- Free — $0. You get GPT-6 Luna at Standard speed in the desktop app, which is still rolling out.
- Go — $8/month. OpenAI lists the same Codex access as Free: GPT-6 Luna at Standard speed in the desktop app.
- Plus — $20/month. Codex works on the web, in the CLI, in the IDE and on iOS, with GPT-6.1 Sol and GPT-6 Luna. OpenAI estimates 15–160 local messages every 5 hours on GPT-6.1 Sol, or 5–45 on GPT-6 Astra. Weekly limits may also apply.
- Pro — $100, $200 or $500/month. According to OpenAI, Pro currently has no five-hour limit. Only the $500 tier gets Astra Ultrafast.
- Business — $25 per user/month on monthly billing. Teams of 2+ can pay less on an annual plan, but the Codex pricing page doesn't give that figure.
Anthropic Claude Code (comes with Claude plans)
- Pro — $20/month, or $17/month if you pay $200 up front for a year. You get at least 5x Free's usage per 5-hour session, plus weekly limits.
- Max 5x — from $100/month, with 5x Pro's usage per 5-hour session.
- Max 20x — 20x Pro's usage. Anthropic doesn't publish the price; the pricing page only says "From $100".
- Team — Standard seats are $25/month ($20 billed annually). Premium seats are $125/month ($100 annually) and come with 5x Standard usage.
Google (Antigravity, Jules)
- Gemini CLI is gone for individual accounts — since June 18, 2026, Gemini CLI no longer serves Google AI Pro, AI Ultra or free individual accounts, and Google is pointing everyone to Antigravity CLI instead. The only people still on Gemini CLI are enterprise users with Gemini Code Assist licenses and anyone signing in with an API key. Antigravity itself has a free tier with a basic weekly allowance; Google doesn't say how big it is.
- Google AI Plus — $4.99/month (US). Google's plan page doesn't list Antigravity as a Plus perk, and in Antigravity any account that isn't on Pro or Ultra, Plus included, gets a quota that only refreshes weekly.
- Google AI Pro — $19.99/month (US) or $199.99/year. Your Antigravity quota, CLI included, refreshes every five hours until you hit a weekly limit, and you can buy AI credits to keep going after that. Limits also go up in Jules and AI Studio.
- Google AI Ultra — $99.99/month (5x Pro usage) or $199.99/month (20x), billed monthly only. In Antigravity you get the biggest quota, also refreshed every five hours, the highest weekly limit and access to third-party models. Google hasn't published the actual numbers.
Cursor
- Free — $0. Limited agent and chat requests.
- Plus — $8/month. Shared AI usage across Cursor and Grok; Grok models only in Cursor on this tier.
- Premium — $30/month (4x Plus usage). Cloud agents and automations, Origin code hosting, Bugbot reviews, and third-party models at API list price (billed separately).
- Super — $100/month (15x Plus). Priority model access at peak times.
- Ultra — $200/month (50x Plus). Earliest access to products and models.
- Teams — Premium $30, Super $100, or Ultra $200 per seat per month; seats can mix tiers.
GitHub Copilot
- Free — 2,000 completions a month, with limited chat and agent use.
- Pro — $10/month for 1,500 AI credits (1,000 base plus 500 flex). One credit is worth $0.01.
- Pro+ — $39/month, 7,000 credits.
- Max — $100/month, 20,000 credits.
- Business is $19 per seat (1,900 credits) and Enterprise $39 per seat (3,900 credits).
- On paid plans, code completions and next-edit suggestions don't cost credits. GitHub says the flex portion could change over time.
Chinese coding plans (international versions)
- Alibaba Cloud Qwen Coding Plan Pro — $50/month, capped at 6,000 requests per 5 hours, 45,000 a week and 90,000 a month. It covers Qwen, GLM, Kimi and MiniMax models. Slots are limited.
- Z.ai GLM Coding Plan — from $18/month. Lite gives 2,000 credits per 5 hours and 10,000 a week. The docs spell out the Pro (12,000/60,000) and Max (28,000/140,000) allowances but not their prices.
- MiniMax M Plan — Go $22, Explore $55 (3x Go's usage), Build $132 (7.5x).
- BytePlus ModelArk Coding Plan — Lite $10/month, Pro $50/month.
- StepFun Step Plan — four tiers at $6.99, $9.99, $29 and $99/month.
- Kimi membership (includes Kimi Code) — ¥49, ¥99, ¥199 or ¥699/month, priced in yuan only.
- Tencent Hy Token Plan — ¥28, ¥78, ¥238 or ¥468/month, for mainland China.
Entry tier (under $25)
You don't have to pay anything to get started. Codex on ChatGPT Free, Copilot Free and Cursor Free all cost nothing, and Google's Antigravity has a free tier too, though its allowance only refreshes weekly.
The cheapest paid options are Step Plan Flash Mini ($6.99), ChatGPT Go ($8), Copilot Pro ($10) and BytePlus Coding Plan Lite ($10). One caveat: Go and Free only get GPT-6 Luna in the desktop app. To use GPT-6.1 Sol in the CLI, the IDE or the cloud, you need Plus.
Most individual developers start around $20, with ChatGPT Plus, Claude Pro or Google AI Pro. GLM Lite is a little cheaper at $18, and MiniMax Go a little more at $22. Each vendor counts usage differently:
- Claude and Codex use a rolling 5-hour window, with a weekly cap layered on top.
- Google's Antigravity works along similar lines: Pro and Ultra refresh every five hours under a weekly cap, while everyone else gets a weekly allowance.
- Cursor and Copilot just tell you how many dollars or credits of usage the plan includes.
Heavy-use tier ($100+)
If you need more, $100 is the usual next step: ChatGPT Pro $100, Claude Max 5x, Copilot Max (20,000 credits, or $200 of usage at face value) and Google AI Ultra 5x at $99.99.
Above that sit Cursor Ultra and ChatGPT Pro $200, Google AI Ultra 20x at $199.99, and Claude Max 20x, whose price isn't published. At the top is ChatGPT Pro $500, the only individual plan with GPT-6 Astra Ultrafast.
Two Chinese plans land in this range: MiniMax Build ($132) and Step Plan Flash Max ($99).
Chinese coding plans: where to buy and how they're metered
These plans sell you usage, not a tool. You get a dedicated API key and base URL and plug them into Claude Code, Cursor, Codex, OpenCode, Cline or something similar.
- Qwen (Alibaba Cloud Model Studio, international) — Pro is the only plan on sale, at $50. It's meant for interactive use in coding tools; running the key from scripts or backends breaks the rules. New slots open daily at 00:00 UTC+8.
- GLM (Z.ai) — priced in USD from $18. Outside peak hours (Monday–Friday, 14:00–18:00 UTC+8), credits go half as fast. From September 25 to October 7, Z.ai charged off-peak rates all day.
- MiniMax (international) — every M Plan tier includes the M3.1 Flash Preview text model, but only Explore and Build add the H3 video model. Usage resets on a 5-hour window and a weekly window.
- ByteDance — BytePlus (international) comes in $10 and $50 tiers; the Volcengine Ark version for mainland China costs ¥40 or ¥200. BytePlus says Lite offers about 3x Claude Pro's quota and Pro offers 5x Lite.
- StepFun — the international Step Plan is priced in USD and tops up credits monthly (400M to 40,000M; roughly 7M credits ≈ $1). The mainland China version runs ¥49–¥699.
- Kimi and Tencent Hy — neither publishes a USD price. Kimi Code draws on the same credit pool as the rest of the membership and also has a separate limit of its own.
Recent changes
- Cursor moves to The X Subscription — on October 13, cursor.com/pricing lists Free, Plus ($8), Premium ($30), Super ($100) and Ultra ($200), with Teams from $30/seat. Usage is a shared weekly pool across Cursor and Grok; Plus is Grok-only in Cursor, and third-party models stay pay-as-you-go from Premium up. Existing Cursor plans can keep running; switching is optional per xAI's docs.
- OpenCode Go raises its DeepSeek V4.1 Flash allowance — on October 12, OpenCode lifted the monthly usage for DeepSeek V4.1 Flash on the $10 Go plan from $60 to $80, and Go Plus at $40 now gets $160. The Go page already shows the new numbers. OpenCode didn't say whether the bump is temporary or permanent.
- Qwen Coding Plan Lite is gone — new subscriptions stopped on March 20, 2026, and renewals and upgrades on April 13. Pro at $50 is all that's left.
- MiniMax switched to M Plan — new buyers now get M Plan instead of Token Plan. The first month on monthly billing is half price until October 14.
- ChatGPT Pro now has three tiers — $100, $200 and $500.
- GPT-5.5 retires on October 14, 2026 — after that you won't find it in ChatGPT, ChatGPT Work or Codex on any plan. The API, OpenAI says, is unaffected. Our pick for a replacement is GPT-6.1 Sol, which OpenAI's Codex Models page recommends for serious coding work; Free and Go don't get it at launch, so they're pointed to GPT-6 Luna in the desktop app.
- Gemini CLI stops serving individual accounts — as of June 18, 2026, Google AI Pro, AI Ultra and free individual accounts can no longer use Gemini CLI. Google is moving people to Antigravity CLI, which can import your existing skills, MCP servers and agent setups. Enterprise Gemini Code Assist licenses and API-key sign-ins aren't affected.
About these numbers
Before writing this, we went through every vendor's official pricing page on October 8. Some vendors don't spell out their own prices: Claude Max is only listed as "from $100", with no separate price for the 20x tier, and GLM gives allowances for Pro and Max but no prices. When that happened, we left the gap rather than guess. Most usage figures are the vendors' own estimates, and how far they really go depends a lot on which model you pick and what you ask it to do. These plans get repriced almost every month, so we check them again on the 1st of each month.
