雷价通
Plan Framing · Official Rates · Community Signal
Membership Rates
Vendor FramingReal Usage Feel
This page puts OpenAI, Claude, Gemini, Cursor, and other plan systems side by side,
so you can read the official framing first and then layer in community signal for real workload intensity.
Open Rate BoardSee Plan Pricing
vendor rate board

Pricing, benefits, and quota framing are organized from official pages. Community notes are for workload feel only.

Free
Trial
$0.00
Official page prevails
A limited entry tier for testing the product and model access.
  • Limited GPT-5.5 Instant access
  • Capped messages and uploads
  • Limited, slower image generation
  • Limited deep research access
  • Reduced memory and context support
  • Limited Codex usage
The official pricing page positions Free as a capped entry tier with limits across messages, uploads, image generation, deep research, memory, and Codex.
Reviewed: 05/14/2026
Go
Base Boost
$8
Official page prevails
A lighter paid tier that is better for steady chat, uploads, and image use.
  • Everything in Free
  • More GPT-5.5 Instant usage
  • Higher message allowance
  • Higher upload allowance
  • Higher image allowance
  • Longer memory retention
Shown at the current $8 tier, with the main value being higher base allowances than Free.
Reviewed: 05/14/2026
Plus
Core Personal
$20.00
Approx. ¥137 / mo
Best for frequent personal use, lighter local coding, and occasional cloud tasks.
  • Everything in Go
  • GPT-5.5 Thinking for stronger reasoning
  • Higher message and upload quotas
  • More advanced image generation
  • More deep research and agent-mode headroom
  • Projects, tasks, and custom GPTs
  • Higher Codex usage
OpenAI's help center notes that Plus offers roughly double regular Codex usage through 2026-05-31.
Reviewed: 05/14/2026
Pro 5x
Power Personal
$100
Official page prevails
For higher quotas, stronger reasoning, and meaningfully larger Codex capacity.
  • Everything in Plus
  • Roughly 5x overall usage
  • Roughly 10x Codex quota
  • GPT-5.5 Pro reasoning
  • Largest Codex task capacity
  • Unlimited GPT-5.3 and file uploads
  • Unlimited and faster image generation
Displayed at the current $100 Pro 5x tier.
Reviewed: 05/14/2026
Pro 20x
Heavy Personal
$200
Official page prevails
For the heaviest personal workflows, especially parallel work and very long sessions.
  • Everything in Pro 5x
  • Roughly 20x overall usage
  • Roughly 20x Codex quota
  • Better fit for sustained heavy sessions
  • More stable for heavy multi-tasking
Displayed at the current $200 Pro 20x tier.
Reviewed: 05/14/2026
Business
Teams
$25 / seat
Official page prevails
A secure shared workspace for collaboration, connected apps, and admin control.
  • ChatGPT Plus plus Business Codex capabilities
  • Top-tier models for work use
  • 60+ app connectors
  • Shared projects and workspace GPTs
  • Member, role, and billing management
  • Controls like SAML SSO and MFA
  • Business data excluded from training by default
Shown at $25 per seat; the enterprise Codex docs also expose the credits-based billing view.
Reviewed: 05/14/2026
Enterprise
Large Orgs
Contact Sales
Official page prevails
Enterprise-grade security, controls, support, and custom commercial terms.
  • Expanded context window
  • SCIM, EKM, analytics, and domain verification
  • Custom data retention policies
  • Encryption at rest and in transit
  • Data residency across 10 regions
  • 24/7 priority support and SLA
Best for organizations that need data residency, compliance features, SLAs, invoicing, and volume discounts.
Reviewed: 05/14/2026
Official Sources:Codex Pricing|ChatGPT Pricing|Codex Help Center

Personal Tier Matrix

OpenAI Official RangeCommunity Signal

The official table is a range, not an exact task count. Community notes help more when judging whether heavy tasks will burn out early.

Task TypeWindowPlus$100$200Community Signal
GPT-5.5 local codingPer 5 hours15 - 80 次80 - 400 次300 - 1600 次Light tasks land closer to the top of the range; multi-file refactors, long context, and higher reasoning push usage toward the lower end.
GPT-5.4 local codingPer 5 hours20 - 100 次100 - 500 次400 - 2000 次Mid-size repo fixes usually last longer than GPT-5.5, but sustained large patches still compress the window quickly.
GPT-5.4 mini local codingPer 5 hours60 - 350 次300 - 1750 次1200 - 7000 次Quick Q&A, comments, and small diffs get closest to the upper bound. It works well as a quota-saving default.
GPT-5.3-Codex local codingPer 5 hours30 - 150 次150 - 750 次600 - 3000 次Many users find it steadier for maintenance work in older repos, but long-chain problems still push it into the lower-middle range.
GPT-5.3-Codex cloud tasksPer week10 - 60 次50 - 300 次200 - 1200 次Cloud tasks are the most sensitive to context size and execution length. Running several long tasks in parallel usually lowers real usable counts.
GPT-5.3-Codex code reviewPer week20 - 50 次100 - 250 次400 - 1000 次Short PR reviews stay closer to the official range; large PRs, cross-directory diffs, and repeated follow-ups burn quota faster.
Community Note
Short tasks stay near the upper bound
Single-file fixes, pre-release checks, and small copy edits usually land closer to OpenAI's theoretical upper range than larger refactors do.
Community Note
Complex repos cut usable counts fast
Multi-file diffs, long context, repeated follow-ups, and high-reasoning mode all reduce how many tasks fit into the same 5-hour window.
Community Note
Cloud tasks and review fluctuate most
These are most affected by context size, execution length, and concurrency, so public samples usually suggest budgeting around the midpoint of the official range.
One Plus user burned ~75% of a 5h window in 20 minutes
Public sample · 04/11/2026

Public examples consistently show that multi-file refactors, long context, and high-reasoning mode can push Plus far below the official upper range.

View Source Thread
Some users observed a roughly 2:1 weekly-to-5h burn ratio
Public sample · 03/11/2026

That implies you can hit the weekly limit before fully exhausting each 5-hour window if the workload stays intense across the week.

View Source Thread
Lighter daily workflows may still leave 10% to 20% weekly headroom
Public sample · 04/12/2026

If most work is short Q&A, local patching, and code review, Plus or lower Pro tiers feel much better than they do in heavy-repo scenarios.

View Source Thread