AI coding tools have transformed from experimental sidekicks into core development infrastructure. Five global leaders : GitHub Copilot, Cursor, Claude Code (Anthropic), Codex CLI (OpenAI), and DeepSeek Harness : now compete alongside three major Chinese AI coding assistants: Tongyi Lingma (Alibaba Cloud), CodeGeeX (Zhipu AI / Tsinghua University), and Baidu Comate.
Checked against vendor pages on 2026-10-27. Where a vendor shows only an annual rate, this page shows that rate and says so.
Sources: github.com/features/copilot/plans, cursor.com/pricing, docs.github.com/en/copilot, claude.ai/pricing, tongyi.aliyun.com/lingma/pricing, codegeex.cn, cloud.baidu.com/doc/COMATE/s/rlnvnio4a
Pricing Comparison at a Glance
| Plan | GitHub Copilot | Cursor | Claude Code | Tongyi Lingma | CodeGeeX | Baidu Comate |
|---|---|---|---|---|---|---|
| Free | Free | Free, limited Agent and Tab | Free (chat only, no Claude Code) | Personal Basic: Free | Free (all core features, unlimited) | Personal Standard: Free |
| Entry Pro | Pro: $10/mo | Plus: $8/mo | Pro: $20/mo ($17/mo annual) | Personal Pro: ¥59/mo (~$8.7/mo) | — (personal use is free) | Personal Pro: ¥100/mo (~$14.7/mo) |
| Mid tier | Pro+: $39/mo | Premium: $30/mo | Max 5x: $100/mo | Enterprise Standard: ¥99/seat/mo (~$14.6/mo, 10+ seats) | Enterprise: custom pricing | Enterprise Pro: ¥150/user/mo (~$22/mo) |
| High tier | Max: $100/mo | Super: $100/mo / Ultra: $200/mo | Max 20x: $200/mo | Enterprise Dedicated: ¥199/seat/mo (~$29.3/mo, VPC, 100+ seats) | Private deployment: custom pricing | Enterprise Dedicated: ¥2,500/user/yr (~$35/mo, 100+ seats) |
| Teams | Business: $19/user/mo | Teams: $30+/user/mo | Team Premium: $125/seat/mo | — | — | — |
| Enterprise | Enterprise: $39/user/mo | Custom | Custom | Enterprise Dedicated: ¥199/seat/mo | Custom private deployment | Enterprise Dedicated: ¥2,500/user/yr |
| Student discount | Free for verified students | Free Pro for 1 year (.edu) | Claude for Open Source: 6 mo Max 20x | Free for all | Free for all | Free tier available |
| Usage billing | Usage-based AI Credits (since June 1, 2026) | Monthly credit pool per plan | Token budget resets every 5 hours | — | — | Agent request packs (pay-as-you-go) |
Source: GitHub Copilot plans | Cursor pricing sources | Tongyi Lingma pricing | CodeGeeX | Baidu Comate pricing
Which prices this page verified
The Copilot, Cursor, Claude Code, and Codex figures come from those vendors' own pages on 27 October 2026.
The three Chinese tools need a separate check. The Tongyi Lingma pricing page now lists tiers at ¥0, ¥59, ¥169, and ¥559, which does not match the earlier per-seat structure. The CodeGeeX site and the Baidu Comate documentation returned no published price. The Chinese-tool figures in the table above are therefore not verified, and this page does not rest on them. Check each vendor page before you buy.
Key Pricing Takeaways
- GitHub Copilot Pro at $10/month is the cheapest entry. GitHub now bills in AI credits, where 1 credit costs $0.01. Pro includes $15 of monthly credits, Pro+ includes $70, and Max includes $200.
- Cursor replaced its $20 Pro plan. Plus costs $8/mo, Premium costs $30/mo, Super costs $100/mo, and Ultra costs $200/mo.
- Claude Code at $20/month gives you Pro access with a token budget that resets every 5 hours. Heavy users will need Max 5x ($100/mo) or Max 20x ($200/mo).
- Tongyi Lingma now lists tiers at ¥0, ¥59, ¥169, and ¥559 on its pricing page. The earlier per-seat structure no longer appears there.
- CodeGeeX is completely free for individual developers : all core features, unlimited usage, no subscription required. Enterprise private deployment is the only paid offering.
- Baidu Comate publishes no price on its documentation page. Ask Baidu Cloud for a quote before you plan a budget.
- Business buyers: Copilot Business ($19/user/mo) is much cheaper than Cursor Teams ($40/user/mo) and Claude Team Premium ($125/seat/mo). Chinese tools offer competitive enterprise pricing starting at ¥99/seat/mo (Tongyi Lingma).
- For cost-conscious developers: CodeGeeX (free) and Tongyi Lingma Personal Pro (¥59/mo) are unbeatable value.
Billing Model Comparison
A per-plan model list needs a vendor page that states it. Neither the Copilot plans page nor the Cursor pricing page publishes one, so this table covers billing instead. That number is published.
| Billing unit | GitHub Copilot | Cursor | Claude Code |
|---|---|---|---|
| Metered unit | AI credits, 1 credit = $0.01 | Plan credit pool | Token budget per 5-hour window |
| Pro monthly allowance | $15 of credits | Plan dependent | Pro usage limit |
| Top-tier allowance | $200 of credits on Max | Ultra at $200/mo | Max 20x, 20x Pro usage |
| Extra usage | Credits beyond the allowance | Plan dependent | Upgrade to Max 5x or Max 20x |
Sources: github.com/features/copilot/plans, cursor.com/pricing, anthropic.com/pricing
GitHub Copilot: Best for VS Code and JetBrains Users
Pricing: Free, Pro $10/mo, Pro+ $39/mo, Max $100/user/mo, Business $19/user/mo, Enterprise $39/user/mo
What it is: GitHub Copilot is the incumbent. It integrates directly into VS Code, JetBrains, and other major IDEs. In 2026, it goes far beyond code completions: Copilot Chat, Copilot Agent, Copilot Code Review, and Copilot CLI are all included.
Strengths:
- Lowest Pro price: $10/month is hard to beat. Free tier is generous enough for hobbyists.
- Deep GitHub integration: Copilot pulls context from your repos, issues, and PRs. The cloud agent can automate entire workflows.
- Multi-model access: GitHub does not publish a per-plan model list. The plans page states the credit allowance for each tier, so compare tiers on credits, not on model names.
- Code review built in: Automated PR reviews powered by AI. This alone saves teams hours per week.
- Copilot CLI: Natural language shell commands. Describe what you want in plain English.
Weaknesses:
- Usage-based billing: Since June 1, 2026, Copilot uses AI Credits. Heavy usage on premium models will cost extra.
- Pro+ at $39/mo: It adds $70 of monthly credits instead of $15, and Max adds $200. GitHub bills in AI credits at $0.01 each.
- Editor experience varies: Works best in VS Code. JetBrains and other IDEs lag behind in features.
Best for: Developers already in the GitHub ecosystem. The $10/mo Pro plan is the best value if your workflow stays within the premium-request allowance.
Source: github.com/features/copilot/plans
Cursor: Best AI-Native Code Editor
Pricing: Free, Plus $8/mo, Premium $30/mo, Super $100/mo, Ultra $200/mo, Teams $30+/user/mo
What it is: Cursor is a fork of VS Code rebuilt from the ground up for AI-assisted coding. Tab completions are near-instant, and the Agent mode can read your entire codebase, make multi-file edits, and run terminal commands autonomously.
Strengths:
- Best Tab completions in class: Cursor predicts your next edit with uncanny accuracy, delivering multi-line, context-aware suggestions.
- Agent mode: Cursor's agent can read your whole project, create files, run commands, and fix bugs autonomously. This is the most capable coding agent on the market today.
.cursorrules: Project-level instructions that shape every completion and agent behavior. This is a significant advantage for team consistency.- Generous free tier: Hobby gives you limited Agent mode without a credit card.
- Credit system that rarely hits you: Auto mode uses your local model effectively for free. Credits only deplete on premium models in Max mode.
Weaknesses:
- $20/mo entry: Pro is double Copilot Pro's price. For light users, Copilot delivers more value.
- Not a standalone product: It's a fork of VS Code. If your team is on JetBrains or another IDE, Cursor requires migration.
- Teams pricing is steep: $40/user/mo is more than double Copilot Business ($19/user/mo).
Best for: Developers who want the best possible AI coding experience and are willing to pay for it. The Agent mode alone justifies the $20/mo for professional developers.
Claude Code: Best for Agentic Workflows
Pricing: Free (chat only), Pro $20/mo, Max 5x $100/mo, Max 20x $200/mo, Team Premium $125/seat/mo
Claude Code is Anthropic's terminal-native coding agent. It's a CLI tool that runs in your terminal, reads your codebase, and executes complex multi-step development tasks.
Strengths:
- Best-in-class agent execution: Claude Code handles complex, multi-file refactoring better than any competitor. It writes the code changes, not just suggestions.
- Terminal-first workflow: No IDE dependency. Works with any editor (VS Code, Neovim, JetBrains) equally well.
- Opus reasoning: Claude Opus 5.5 is the current top model in the family. It handles tasks that need deep logical reasoning.
- Context handling: Claude Code keeps coherence across very long contexts. It suits large codebases.
- Open source maintainer program: 6 months of Max 20x free for qualifying projects.
Weaknesses:
- Agent-first, not editor-first: You interact through the terminal, not inline in your editor. Developers who want inline completions will prefer Cursor or Copilot.
- Token budget limits: Pro's budget resets every 5 hours. Heavy daily users hit this ceiling and must upgrade to Max plans ($100-200/mo).
- No free coding tier: The free Claude plan excludes Claude Code. You must pay at least $20/mo.
- Team pricing is expensive: Team Premium at $125/seat/mo is much more than competitors.
Best for: Developers doing complex refactoring, large codebase work, and agentic workflows. The tool for developers who need AI to execute complex work autonomously.
Codex CLI: Best for OpenAI Model Users
Pricing: Codex is included on every ChatGPT plan, including Free and Go. Codex Cloud needs Plus or higher. ChatGPT plans cost Free $0, Go $8/mo, Plus $20/mo, and Pro $200/mo.
Codex CLI is OpenAI's terminal-native coding agent. Similar to Claude Code but uses GPT models. Has shell access. Can edit files, run commands, debug issues. Integrates with GitHub.
Strengths:
- OpenAI model quality
- Shell access
- GitHub integration
- Configurable providers (via config.toml)
- 400K context window
Weaknesses:
- Smaller community than Claude Code
- Less mature ecosystem
- Pricing can escalate on heavy usage
Best for: Developers already in the OpenAI ecosystem who want a terminal agent with GPT model quality.
Source: github.com/openai/codex
DeepSeek Harness: Best Open-Source Value
Pricing: Free (self-hosted). API costs only: DeepSeek-V4 Flash $0.14 input / $0.28 output per million tokens. V4 Pro $0.435 / $0.87. 10-100x cheaper than OpenAI/Anthropic.
DeepSeek Harness (dsh) is an open-source agent harness built on an everything-is-a-plugin architecture. Uses DeepSeek models. Can run locally or on cloud. Flexible architecture. Good for cost-sensitive projects.
Strengths:
- Open source
- DeepSeek models (10-100x cheaper)
- Local deployment option
- Flexible plugin architecture
- No vendor lock-in
Weaknesses:
- Smaller community
- Less polished UX
- Requires setup
- Model quality slightly below GPT-4/Claude
Best for: Cost-sensitive developers and startups. Self-hosted option for privacy-sensitive projects.
Source: github.com/deepseek-ai/deepseek-harness
Tongyi Lingma: Best Value Chinese AI Coding Assistant
Pricing: Personal Basic Free, Personal Pro ¥59/mo (~$8.7/mo), Enterprise Standard ¥99/seat/mo (10+ seats), Enterprise Dedicated ¥199/seat/mo (VPC, 100+ seats)
Tongyi Lingma (通义灵码) — renamed Qoder CN on May 20, 2026 — is Alibaba Cloud's AI coding assistant powered by the Tongyi Qwen LLM. It integrates with VS Code, JetBrains, and Visual Studio, offering real-time code completion, code generation from comments, unit test generation, and smart Q&A. A standout feature is its cross-file awareness : it understands context across files in your project for more accurate suggestions. The AI Programmer (Beta) mode supports multi-file batch modifications and test generation.
Strengths:
- Aggressively priced: Personal Pro at ¥59/mo (~$8.7) is the cheapest paid plan of any tool here. Its limited-time free promotion ended May 20, 2026.
- Solid feature parity: Code completion, chat, code review, unit test generation, code explanation, and comment generation are all included even on the free tier.
- AI Programmer (Beta): Multi-file editing agent that can batch-modify code and generate unit tests across your project.
- Enterprise features: Enterprise plans include knowledge base Q&A, VPC deployment, audit logging, and custom model inference.
- Chinese language support: Native Chinese comprehension for Q&A and documentation generation.
Weaknesses:
- No free Agent mode: The AI Programmer is limited to 50 conversation rounds and 50 test methods per month on the free plan.
- Less known outside China: The ecosystem and community are primarily Chinese-language.
- No standalone IDE: It's a plugin, not a forked editor like Cursor.
Best for: Chinese-speaking developers and teams using Alibaba Cloud infrastructure. At ¥59/mo, Personal Pro remains the cheapest paid plan of any tool here. Enterprise teams get strong compliance features with VPC deployment and audit logging.
CodeGeeX: Best Free Chinese AI Coding Assistant
Pricing: Free for individual developers (all core features, unlimited usage), Enterprise private deployment (custom pricing)
CodeGeeX is developed by Zhipu AI in collaboration with Tsinghua University. It uses the open-source CodeGeeX4-ALL-9B model with a 128K context window. It supports 300+ programming languages across VS Code and JetBrains IDEs. CodeGeeX offers code generation, completion, translation between languages, smart Q&A, code commenting, and unit test generation. One distinctive feature is Function Call capability with a claimed 90% success rate. It lets the assistant call toolchains directly instead of only generating code.
Strengths:
- Completely free for individuals: No usage limits, no feature gating, no subscription needed. This is the most generous free tier of any AI coding assistant on the market.
- Open-source model: CodeGeeX4-ALL-9B is open source. It runs on a single GPU with 24GB minimum, so you can deploy it locally for private projects.
- 128K context window: Handles large codebases and project-level context better than many competitors.
- Function Call capability: Over 90% success rate on function calls : AI can actually invoke APIs and toolchains, not just generate code snippets.
- Code translation: Translate code between 15+ programming languages, useful for migration projects.
Weaknesses:
- No Agent mode: CodeGeeX focuses on code completion, generation, and Q&A. It lacks the multi-file agent capabilities of Cursor or Claude Code.
- No enterprise tier on website: Enterprise pricing requires contacting sales. No published self-serve business plans.
- Smaller ecosystem: Fewer community extensions and third-party integrations compared to Copilot or Cursor.
Best for: Individual developers who want a powerful AI coding assistant without paying anything. Also great for teams that need open-source models for local/private deployment. Students and hobbyists get the most value here.
Baidu Comate: Best for Spec-Driven and Agentic Development
Pricing: Personal Standard Free, Personal Pro ¥100/user/mo ($14.7/mo), Enterprise Pro ¥150/user/mo ($22/mo), Enterprise Dedicated ¥2,500/user/yr (~$35/mo, 100+ seats)
Baidu Comate (文心快码, aka Wenxin Kuaijian) is Baidu's AI coding assistant powered by the ERNIE-Code model. It's the most feature-rich Chinese offering, supporting VS Code, JetBrains, Visual Studio, Xcode, and Eclipse. Comate features code completion, Zulu Agent (end-to-end task execution), Spec mode (spec-driven development), Figma-to-code conversion, code review, MCP support, and a customizable Skills/Rules/Memory system. Its Comate AI IDE is a standalone editor built for AI-native development.
Strengths:
- Zulu Agent: End-to-end agent that autonomously plans, implements, and validates code from natural language requirements. Similar in capability to Cursor's Agent mode.
- Spec mode: Write specifications in natural language and Comate generates the implementation : unique among Chinese tools.
- Broad IDE support: Supports Xcode and Eclipse in addition to VS Code, JetBrains, and Visual Studio : the widest IDE coverage of any Chinese tool.
- MCP support: Compatible with the Model Context Protocol for extending agent capabilities.
- Comate AI IDE: A standalone AI-native IDE (similar to Cursor) with preview, element-level debugging, and Rule/Memory systems.
- Figma-to-code: Convert design mockups directly into code.
Weaknesses:
- Paid Personal Pro: ¥100/mo (~$14.7) is cheaper than Cursor ($20) but more than Tongyi Lingma's ¥59/mo. The free tier has limited agent capabilities.
- Agent usage is metered: The Zulu agent consumes tokens from request packs. Heavy users must buy additional packs.
- Chinese-centric: Documentation and community primarily in Chinese. Less accessible for English-speaking developers.
Best for: Development teams using Baidu Cloud or needing broad IDE support (Xcode, Eclipse). The Spec mode is ideal for teams that want spec-driven development workflows. The Zulu agent competes directly with Cursor's Agent mode at a lower price point.
Source: Baidu Comate pricing | Baidu Comate features
Feature Comparison Table
| Feature | Copilot Pro ($10) | Cursor Pro ($20) | Claude Code Pro ($20) | Tongyi Lingma (¥59/mo) | CodeGeeX (Free) | Baidu Comate (¥100/mo) |
|---|---|---|---|---|---|---|
| Inline completions | ✅ Unlimited | ✅ Best in class | ❌ (terminal CLI) | ✅ Unlimited | ✅ Unlimited | ✅ Unlimited |
| Chat | ✅ In-editor | ✅ In-editor | ✅ Terminal-native | ✅ In-editor | ✅ In-editor | ✅ In-editor |
| Multi-file Agent | ✅ Copilot Agent | ✅ Agent mode | ✅ Best in class | ✅ AI Programmer (Beta) | ❌ | ✅ Zulu Agent |
| Code review | ✅ Built-in | ❌ (via extension) | ❌ | ✅ Built-in | ❌ | ✅ Built-in |
| CI/CD integration | ✅ GitHub Actions | ❌ | ❌ | ❌ | ❌ | ❌ |
| Terminal commands | ✅ Copilot CLI | ✅ Built-in | ✅ Native | ❌ | ❌ | ❌ |
| Project context | GitHub context | Full codebase index | Full codebase index | Cross-file awareness | 128K context window | Full codebase + Skills/Rules |
| IDE support | VS Code, JetBrains, etc. | VS Code fork | Any (CLI tool) | VS Code, JetBrains, VS | VS Code, JetBrains | VS Code, JetBrains, VS, Xcode, Eclipse |
| Model | GitHub does not publish per-plan models | Claude, GPT, custom | Claude Opus/Sonnet | Tongyi Qwen | CodeGeeX4-ALL-9B | ERNIE-Code |
| Open source | ❌ | ❌ | ❌ | ❌ | ✅ Model open source | ❌ |
| Best for | GitHub ecosystem | AI-native editing | Complex agent tasks | Budget/value | Free + open source | Broad IDE + Agent |
Quick Decision Guide
| If you... | Choose | Starting Price |
|---|---|---|
| Use VS Code + GitHub daily | Copilot Pro | $10/mo |
| Want the fastest Tab completions | Cursor Pro | $20/mo |
| Need complex multi-file refactoring | Claude Code Pro | $20/mo |
| Want the cheapest paid plan | Tongyi Lingma Personal Pro | ¥59/mo (~$8.7) |
| Want a completely free tool | CodeGeeX | $0 |
| Need Xcode or Eclipse support | Baidu Comate | ¥100/mo (~$14.7) |
| Are on a tight budget | CodeGeeX or Copilot Free | $0 |
| Lead a dev team of 10+ | Copilot Business ($19) or Tongyi Lingma Enterprise (¥99) | $19/mo or ~$14.6/mo |
| Are a student | CodeGeeX or Tongyi Lingma (free) or Copilot Free | $0 |
| Want open-source/local deployment | CodeGeeX | $0 |
Summary
The AI coding tool market in 2026 offers eight strong contenders across five global and three Chinese products:
- GitHub Copilot ($10/mo): The most affordable Western pro plan and best-integrated for VS Code + GitHub users. Usage-based billing is the catch: heavy usage on premium models adds up.
- Cursor ($20/mo): The best all-around AI coding experience. Superior Tab completions and Agent mode make it the top choice for professional developers who want the best tools.
- Claude Code ($20/mo): The best agentic workflow tool. Unmatched for complex tasks, but not a replacement for inline completions.
- Codex CLI ($8-100/mo): OpenAI's terminal agent. Strong for OpenAI ecosystem users who want GPT model quality with shell access.
- DeepSeek Harness (free): The open-source value leader. 10-100x cheaper API costs and self-hosted deployment for cost-sensitive projects.
- Tongyi Lingma (¥59/mo): The cheapest paid plan overall. Strong enterprise features with VPC deployment and audit logging.
- CodeGeeX (Free): The only completely free AI coding assistant with no usage limits. Open-source model enables local deployment. Ideal for budget-constrained developers.
- Baidu Comate (¥100/mo): The most feature-rich Chinese tool with Zulu Agent, Spec mode, Figma-to-code, and the widest IDE support including Xcode and Eclipse.
For most professional developers, Cursor Premium at $30/mo balances completion quality, agent capability, and predictable pricing. Cursor Plus at $8/mo is the cheapest way to start. If budget is the priority and you work inside GitHub, Copilot Pro at $10/mo is strong value. For zero cost, CodeGeeX gives solid code completion. Teams that need administration compare Copilot Business at $19/user/mo with Cursor Teams at $30+/user/mo.
Pricing sourced from official websites as of May 2026. Check each tool's pricing page for the most current rates. Copilot migrated to usage-based AI Credits billing on June 1, 2026. CNY to USD conversion uses approximate rate of ~¥6.8 = $1.
What changed in this revision
| Claim in the May version | What the vendor pages show now |
|---|---|
| Cursor Pro $20/mo, Pro+ $60/mo, Teams $40/user/mo | Plus $8, Premium $30, Super $100, Ultra $200, Teams $30+ |
| GitHub Copilot had no $100 tier | Copilot Max costs $100/user/mo |
| Copilot Pro includes 300 premium requests at $0.04 each | GitHub bills in AI credits at $0.01 each, with $15 on Pro, $70 on Pro+, and $200 on Max |
| Codex needs a paid plan | Codex ships on every ChatGPT plan, including Free and Go |
| GPT-5 mini, Claude Sonnet 4.6, Opus 4.7, Haiku 4.5 | Claude Opus 5.5, Sonnet 5.5, and Haiku 5.5 are current. GitHub publishes no per-plan model list |
| GPT-5.3-Codex token prices | No vendor page states these. Removed |
| Tongyi Lingma ¥99/¥150/¥199 per seat | The page now shows ¥0, ¥59, ¥169, ¥559. The per-seat structure is gone |
| Baidu Comate Personal Pro ¥100/mo | No price on the Baidu documentation page. Withheld |
The model-access table became a billing-model table. A per-plan model list needs a vendor page that states it, and neither Copilot nor Cursor publishes one.
Try Cursor Pro Free
AI-first code editor with unlimited Tab completions. Free for students with a .edu email.
Get Started — from $20/mo