Hardcore Reviews
Hardcore Reviews

AI Coding Plan Comparison: GLM vs Claude Code vs Cursor vs Windsurf vs Codex

A comparison of 5 AI coding plan subscriptions: Zhipu GLM Coding Plan, Anthropic Claude Code (Pro/Max), Cursor Pro, Windsurf Pro, and OpenAI Codex. Includes a plan comparison table (model/price/features/China availability), selection guidance, and 5 FAQs. Prices per official sites; representative comparison, not a personal benchmark.

Published August 11, 20269 min read
<!-- ai-coding-plan-comparison-review | review | AI Coding Plan Comparison: GLM vs Claude Code vs Cursor vs Windsurf vs Codex -->

In 2026, the battleground for AI coding tools has shifted from "the tool itself" to "the subscription plan." A year ago, choosing Cursor or Tongyi Lingma was about the IDE experience; now the real dilemma is: which plan should your 20 dollars a month go to? Cursor Pro, Claude Code, Windsurf Pro, OpenAI Codex, plus Zhipu's GLM Coding Plan-which just had all user quotas fully reset on 8/11-five mainstream AI coding plans, each backed by a completely different set of models, quotas, and ecosystem logic. This review compares one thing only: what your money actually buys.

Scope first. This site has already published two related reviews: the AI Coding IDE Comparison (comparing IDE form factors and features, Cursor, Trae, Qoder, etc.) and the AI Terminal Coding Agent Comparison (comparing CLI agent autonomy, Claude Code, Codex CLI, Aider, etc.). This is the third piece, focused exclusively on subscription plans: not whether the IDE UI is pretty, not whether CLI commands are smooth, just quotas, pricing, included models, and use cases. Below is compiled from each platform's official site and public information, as of 2026-08-11. Closed-source pricing is uniformly "per official site," no invented quota numbers. This is a representative comparison, not a personal full-scale benchmark. This site also published the ZCode upgrade GLM Coding Plan quota reset hotspot (this batch's hotspot piece), the crush coding agent open-source resource (open-source piece), and the AI coding context engineering SOP (how to use after picking a plan). This piece covers the "plan selection" step.

One: Know the Difference: A Coding Plan Is Not a Tool, It's a Subscription

Three easily confused concepts, demarcated first. AI Coding IDE: you download an editor with built-in AI completion and chat; what you compare is editor experience and plugin ecosystem. This site's IDE comparison covers this. AI Coding Agent: an autonomous agent running in the terminal or background; given a task, it modifies files and runs commands itself; what you compare is autonomy and model binding. This site's CLI agent comparison covers this. AI Coding Plan: the subscription tier you pay for monthly; the core question is "for this money, how much quota do I get, which models can I call, and in which tools can I use them." What you compare is subscription value and quota logic. This piece only compares the third category.

Why pull it out separately? Because by 2026, plan pricing has gotten complex enough to be confusing. Some bill by 5-hour rolling windows, some cap by request count, some lock the strongest models behind higher tiers, some tie quotas to their own IDE. The same "20 dollars/month" buys wildly different things on different platforms. So this piece doesn't compare tool features, only the real money inside the plan.

Two: Five-Plan Side-by-Side Comparison

Five representative plans, compared across backing model or tool, subscription price, highlights, and China availability. Closed-source pricing is uniformly "per official site."

PlanBacking model or toolSubscription priceHighlightsChina availability
Zhipu GLM Coding PlanGLM series + ZCodePer official siteDomestic, tied to ZCode, 8/11 full quota reset, multi-tool compatibleAvailable (domestic)
Anthropic Claude CodeClaude Sonnet/OpusPro $20/mo, Max $100/$200/mo (per official site)Strongest agentic depth, code quality gold standardRestricted, needs VPN
Cursor ProCursor IDE (multi-model)~$20/mo (per official site)VS Code ecosystem, multi-model switching, complete IDE experienceRestricted (network)
Windsurf ProWindsurf IDE (Codeium)~$15/mo (per official site)Value, Cascade agent flow, all-in-one IDERestricted (network)
OpenAI CodexGPT seriesWith ChatGPT Plus $20/Pro $200 (per official site)Bundled with ChatGPT, OpenAI ecosystem, CLI + cloudUnavailable

A few takeaways. First, the five plans represent five strategies: GLM bets on domestic availability, Claude Code bets on agentic depth, Cursor bets on IDE ecosystem, Windsurf bets on value, and Codex bets on the OpenAI ecosystem. Second, "subscription price" is just the entry ticket. What truly determines experience is the quota logic: 5-hour rolling windows, request-count caps, credit-point systems, each with different algorithms. Specific quotas are per official site; don't trust secondhand numbers. Third, China availability is a hard dividing line: GLM is the only domestically available option; the other four are all restricted or unavailable.

Three: One by One: Each Plan's Best Range

Zhipu GLM Coding Plan: the only domestic option, tied to ZCode. Zhipu's coding plan, bound to its in-house ZCode tool, backed by the GLM model series. On 8/11 all user quotas were fully reset (covered in this batch's hotspot piece, see the ZCode upgrade hotspot), making it the most-watched domestic plan. Strengths: domestic availability plus compatibility with multiple tools (Cursor, Cline, Roo Code, and other mainstream tools all work) plus MCP tool support. Developers in China need no VPN and no payment hassle. Shortcomings: GLM models still trail Claude/GPT in complex multi-step reasoning, and ZCode's ecosystem maturity lags VS Code-based tools. Best for: developers in China, compliance-friendly use, multi-tool switching. Pricing per official site. China availability: available, domestic.

Anthropic Claude Code: code quality gold standard, deepest agentic capability. Anthropic's plan is delivered via Claude Pro/Max subscriptions, backed by Claude Sonnet/Opus models. Pro tier is $20/month; Max tiers are $100/month and $200/month (per official site), with different tiers corresponding to different usage limits. Strengths: code quality and agentic depth are consensus first-tier. Long context, complex multi-step refactoring, cross-file understanding: Claude's reasoning is the current benchmark. Shortcomings: highest price, closed-source, model not swappable, and requires VPN in China. Best for: developers and teams pursuing top code quality, complex refactoring, with budget to spare. China availability: restricted, API needs VPN.

Cursor Pro: VS Code ecosystem moat, the IDE plan benchmark. Cursor's subscription tier, ~$20/month (per official site), essentially buying quota to call multiple models (Claude, GPT, etc.) inside a modified VS Code. Strengths: complete VS Code plugin ecosystem, flexible multi-model switching, smoothest IDE experience. If you're a VS Code veteran, migration cost is lowest. Shortcomings: quotas get tight under the 5-hour rolling window with heavy use, pricing is opaque, and China network is restricted. Best for: heavy VS Code users, multi-model switching, IDE-workflow developers. China availability: restricted, unstable network.

Windsurf Pro: the value tier, all-in-one IDE. Windsurf's (formerly Codeium) subscription, ~$15/month (per official site), the lowest entry price of the five. Strengths: value plus Cascade agent flow (runs multi-step tasks in one go) plus integrated IDE experience. Shortcomings: model capability and ecosystem trail Cursor and Claude Code, smaller community. Best for: budget-conscious independent developers wanting an all-in-one IDE agent experience. China availability: restricted, unstable network.

OpenAI Codex: bundled with ChatGPT, OpenAI ecosystem. OpenAI's coding capabilities come with ChatGPT Plus ($20/month) and Pro ($200/month) subscriptions (per official site), including Codex CLI and cloud Codex. Strengths: GPT series model capability plus OpenAI ecosystem integration plus dual CLI and cloud form factors. Shortcomings: unavailable in China, model locked to OpenAI, and coding isn't the only selling point of a ChatGPT subscription (quota is shared with chat). Best for: existing ChatGPT subscribers, OpenAI ecosystem users. China availability: unavailable.

Four: Pick by Need

Direct conclusions by your need.

Your needTop pickReason
Usable in China, no VPN or payment hassleZhipu GLM Coding PlanOnly domestic option, 8/11 quota reset, multi-tool compatible
Top code quality, complex refactoringAnthropic Claude CodeClaude model is the code quality gold standard, deepest agentic capability
Heavy VS Code user, multi-model switchingCursor ProEcosystem moat, lowest migration cost
Budget-conscious, want all-in-one IDE agentWindsurf ProLowest entry price, Cascade agent flow
Already have ChatGPT subscription, OpenAI ecosystemOpenAI CodexBundled with subscription, CLI + cloud dual form
China-available + data stays localGLM Coding PlanOnly domestic option, no VPN needed

Most developers' actual combo: use GLM Coding Plan as the primary in China (available plus compliant), switch to Claude Code (needs VPN) as backup for tasks it can't handle. Well-funded teams go straight to Claude Code Max. No need to subscribe to all five. Pick by region and budget.

Five: Three Pitfalls

One: Don't take "monthly fee" as the whole story; quota logic is the real cost. The same 20 dollars/month might give you hundreds of requests on one platform, throttle you with a 5-hour rolling window on another, or lock the strongest model behind a higher tier on a third. A single agent task can trigger a dozen-plus model calls; what seems "enough" can be exhausted in half a day. Before subscribing, read the quota rules on the official site. Don't trust secondhand numbers. Quota policies change fast; today's numbers may be void tomorrow.

Two: Most closed-source plans are restricted in China; don't pay first and find out you can't use it. Of the five, only GLM Coding Plan is domestically available. The other four (Claude Code, Cursor, Windsurf, Codex) all require a VPN or are outright unavailable. Developers in China should confirm network and payment work before subscribing. Don't pay for a year only to find you can't connect. This is the most common and most painful pitfall.

Three: Annual subscriptions carry risk; the market changes too fast. AI coding plan policies shifted rapidly in 2026: quota adjustments, billing-model switches, plan discontinuations have all happened. Prefer monthly billing, even if slightly pricier, to keep the freedom to leave anytime. Data in this article is as of 2026-08-11. Before purchasing, always check each platform's official site for the latest terms.

Six: FAQ

Q: What's the difference between an AI coding plan and an AI coding IDE? A: An IDE is the editor you download, compared by UI and plugin ecosystem (this site's IDE comparison covers that). A plan is the subscription tier you pay for monthly, compared by quota, models, and value (this piece covers that). Many IDEs (like Cursor, Windsurf) are free to download. What you pay for is the quota to call models inside them. That quota is the plan.

Q: Which ones can developers in China use? A: Of the five, only Zhipu GLM Coding Plan is domestically available: no VPN, no payment hassle, and quotas were just fully reset on 8/11. The other four (Claude Code, Cursor, Windsurf, Codex) all require a VPN or are outright unavailable. Developers in China should look at GLM first, then use Claude Code as backup for complex tasks.

Q: What's the difference between Claude Code Pro and Max? A: Pro is ~$20/month; Max has $100/month and $200/month tiers (per official site). The difference is mainly in usage limits. Higher tiers mean more calls within the 5-hour rolling window and access to stronger models. Specific quotas per official site. Heavy users should go Max; light users are fine with Pro.

Q: Cursor and Windsurf are both IDE plans; how do I choose? A: Cursor is ~$20/month, with the most complete VS Code ecosystem and the most flexible multi-model switching, suited for VS Code veterans. Windsurf is ~$15/month, with a lower entry price and a smooth Cascade agent flow, suited for budget-conscious independent developers. Both are closed-source with restricted China networks. For ecosystem pick Cursor; for value pick Windsurf.

Q: Can I skip the plan and configure my own API key? A: Yes. Open-source tools like Aider and OpenHands are free; you plug in any model's API key and pay per use (this site's CLI agent comparison has details). But plans offer predictable quotas, no need to calculate token costs yourself, and smoother tool integration. Heavy users find plans more cost-effective; light users save more with pay-per-use APIs.


References

This article is AI-assisted and human-edited. Last updated: 2026-08-11

FAQ

What's the difference between an AI coding plan and an AI coding IDE?
An IDE is the editor you download, compared by UI and plugin ecosystem (this site's [IDE comparison](/en/ai-coding-ide-comparison-review) covers that). A plan is the subscription tier you pay for monthly, compared by quota, models, and value (this piece covers that). Many IDEs (like Cursor, Windsurf) are free to download. What you pay for is the quota to call models inside them. That quota is the plan.
Which ones can developers in China use?
Of the five, only Zhipu GLM Coding Plan is domestically available: no VPN, no payment hassle, and quotas were just fully reset on 8/11. The other four (Claude Code, Cursor, Windsurf, Codex) all require a VPN or are outright unavailable. Developers in China should look at GLM first, then use Claude Code as backup for complex tasks.
What's the difference between Claude Code Pro and Max?
Pro is ~$20/month; Max has $100/month and $200/month tiers (per official site). The difference is mainly in usage limits. Higher tiers mean more calls within the 5-hour rolling window and access to stronger models. Specific quotas per official site. Heavy users should go Max; light users are fine with Pro.
Cursor and Windsurf are both IDE plans; how do I choose?
Cursor is ~$20/month, with the most complete VS Code ecosystem and the most flexible multi-model switching, suited for VS Code veterans. Windsurf is ~$15/month, with a lower entry price and a smooth Cascade agent flow, suited for budget-conscious independent developers. Both are closed-source with restricted China networks. For ecosystem pick Cursor; for value pick Windsurf.
Can I skip the plan and configure my own API key?
Yes. Open-source tools like Aider and OpenHands are free; you plug in any model's API key and pay per use (this site's [CLI agent comparison](/en/ai-coding-agent-comparison-review) has details). But plans offer predictable quotas, no need to calculate token costs yourself, and smoother tool integration. Heavy users find plans more cost-effective; light users save more with pay-per-use APIs.

Related

Hardcore Reviews

One compromised agent loses everything: a comparison of four credential and permission governance approaches

Credentials went from a config item to an attack surface, yet most teams' defenses are still stuck at "put the agent in a sandbox." This review splits cleanly from our sandbox-isolation comparison: the sandbox governs where code runs; credential governance governs how secrets are used, who approves actions, and whether they can leave. It contrasts four approaches — OpenClaw 2.0, OpenWorker, OpenHuman and traditional secret storage — across six lifecycle stages (store / use / approve / exfiltrate / audit / multi-agent): OpenClaw with masked requests plus an opt-in proxy allowlist; OpenWorker with hard floors, an autonomy ladder, a reviewer model and a circuit breaker, and never self-approving unattended; OpenHuman with Privacy Mode enforced in the Rust core and E2E-encrypted inter-agent comms. Secondhand data (SaaS Sentinel transcription, no primary source located) shows compromise probability 0.24 with one agent rising to 0.86 with seven — risk grows superlinearly with count, under the premise "any agent proposes, execute."

Sep 1, 202611 min read
Hardcore Reviews

Comparing 11 Models by Real Token Cost After the August 31 Repricing: Peak Hours, Cache Hits, and Tokenizer Effects

A model's list price wears at least three more layers. Time of day: DeepSeek moved to peak and off-peak pricing on August 17, charging peak rates on weekdays from 09:00-12:00 and 14:00-18:00, halving them off-peak, and applying off-peak rates all weekend, so the same model costs twice as much at 3pm as at 10pm. Caching: prefix cache hits are billed far below standard input, and the variable sits with your prompt structure rather than the vendor. Tokenization: Sonnet 5 changed tokenizers, so the same input now maps to 1.0x to 1.35x more tokens, and the multiplier floats with content type. This comparison fixes one unit throughout, blended rate equals input plus output divided by two, assuming equal token volumes, as a neutral starting point, then recalculates under three realistic load profiles across 11 models, covering list price, cached input, peak and off-peak, and post-tokenizer position. The finding is not which model is cheapest, it is that no model is cheapest, only cheapest for your particular load: any comparison that ignores input-output ratio, cache hit rate, and content type is comparing list prices, not costs. Chinese model prices come from a page-by-page check of official pricing pages on 2026-08-24, re-confirmed on 08-28; overseas prices from a 2026-08-31 roundup. Conflicts are flagged per line. No live benchmarking was performed.

Aug 31, 20269 min read
Hardcore Reviews

Agent Harness Compared: DeepSeek Harness vs Claude Code vs OpenCode vs Codex - How to Choose Among the Big Four Coding Agents

A four-way coding-agent harness comparison: DeepSeek Harness (101,905★ MIT, everything-is-a-plugin composable foundation) vs Claude Code (141,500★ out-of-box productivity) vs OpenCode (197,583★ model-agnostic daily driver) vs Codex CLI (106,006★ OpenAI ecosystem). Two tables, best-range-per-tool, layered selection, 5 FAQs. Representative comparison, not a hands-on benchmark; stars are API snapshots; per official sources.

Aug 15, 20269 min read