The cheapest AI coding tool is the one whose limit matches how you work. Entry plans cost $8 a month (Cursor Plus) to $20 (Claude Pro, and ChatGPT Plus according to press reports). The top individual tiers of Cursor and Claude cost $200, and press reports put ChatGPT's top tier at $200 too. The tools differ in what runs out first. It can be requests per day, a five-hour window, a weekly pool or a credit balance.
Prices and model names in this article were updated on 31 October 2026.
Windsurf and Supermaven were in the July 2026 version of this article. Their current plans could not be confirmed on an official page on 31 October 2026. This version leaves them out instead of quoting older figures.
Pricing overview
| Tool | Individual plans (per month) | Team and business | Source |
|---|---|---|---|
| Claude Code | Pro $20 ($17 billed annually), Max $100, Ultra $200 | Team Standard $25 per seat ($20 annually), Team Premium $125 ($100 annually); Enterprise $20 per seat billed annually, 20-seat minimum, usage at API rates on top | claude.com/pricing |
| Cursor | Free, Plus $8, Premium $30, Super $100, Ultra $200 | Teams $30, $100 or $200 per seat; Enterprise custom | Cursor pricing |
| OpenAI Codex | Included in ChatGPT Free, Go, Plus and Pro; press reports Plus $20, Pro $100 and Pro $200 | Included in ChatGPT Business and Enterprise | Codex pricing, MacRumors |
| GitHub Copilot | Free, Pro $10, Pro+ $39, Max $100 | Business $19 per seat, Enterprise $39 per seat | Copilot plans |
| Gemini CLI | Free with a Google account (1,000 requests a day); more requests with Google AI Pro or Ultra | Gemini Code Assist Standard and Enterprise (licence seats) | Quota and pricing |
These are list prices. Cursor, Claude and GitHub all have custom Enterprise terms. OpenAI's Codex pricing page names the plans, but its dollar amounts did not render when the page was read on 30 October 2026. The ChatGPT prices therefore come from press coverage of the April 2026 Pro tiers.
What you get at each price point
Claude Code ($20–$200/month)
Claude Code is Anthropic's coding agent; Pro and Max users run it "in their terminal". Anthropic's pricing page says it "is included in all paid plans" and "shares the same usage limits as the rest of your plan". Those limits "reset on a rolling five-hour session window, and paid plans add weekly limits on top." Claude Code defaults to Claude Opus 5.5 on Pro, Max, Team, Enterprise and the Anthropic API.
| Plan | Price | Usage, as Anthropic describes it |
|---|---|---|
| Pro | $20/month | "More usage than Free"; roughly 45 short messages per five hours (help centre estimate) |
| Max | $100/month | "5x Pro usage every 5 hours and every week"; at least 225 messages per five hours |
| Ultra (formerly Max 20x) | $200/month | Pricing page: "10x Pro usage every week", "No 5-hour limit". Help centre: "20x more usage per session than Pro" |
The two official pages disagree on the $200 tier's multiplier, so treat it as unsettled. Anthropic's post on 30 October 2026 confirms the rename and the prices:
We're renaming our Max plans: Max 5x → Max Max 20x → Ultra Max is still $100/month, and Ultra is still $200. If you're a subscriber, there's nothing you need to do. We're also removing the 5-hour limit for verified Ultra accounts, starting today with most existing subscribers. New…
— Claude (@claudeai) October 30, 2026
Past the plan limit, usage credits keep you working "at standard API rates", with a monthly spending cap you set. Claude Code can also run on an API key or a cloud provider, billed per token. For that route, Anthropic publishes an average of "around $13 per developer per active day and $150-250 per developer per month" across enterprise deployments. The Claude Code cost guide covers both routes in detail.
Cursor ($8–$200/month)
Cursor replaced its Hobby, Pro and Pro+ plans with five tiers; the old ones are now listed as legacy plans.
| Plan | Price | Included usage |
|---|---|---|
| Plus | $8 | Base usage; Cursor's own models only |
| Premium | $30 | 4x Plus; third-party models (Claude, GPT, Gemini) at API list price from prepaid credit |
| Super | $100 | 15x Plus; priority model access at peak |
| Ultra | $200 | 50x Plus; earliest access to new products and models |
The Plus plan excludes "Auto, Cloud Agents, Automations, Bugbot, Origin, third-party models, or your own API keys". Cursor's own models (Grok 4.7, 4.6 and 4.5, and Composer 2.5) draw from a weekly pool. Beyond that pool, usage comes from prepaid credit at per-token rates; Grok 4.7 costs $2 per million input tokens and $6 per million output. Auto Top-up and a monthly additional-usage limit control how far that can go.
OpenAI Codex (included in ChatGPT plans)
Codex is OpenAI's coding agent. The July version of this article called it "GitHub Codex", which was wrong. OpenAI's pricing page says Codex is included in "ChatGPT Free, Go, Plus, Pro, Business, or Enterprise" plans.
Local messages and cloud tasks share a five-hour window, and "Additional weekly limits may apply." Plus and Pro users can buy credits. Any user can run extra local tasks with an API key "at standard API rates". Business and new Enterprise plans moved credit use to "API-style token pricing, based on input, cached input, and output tokens" from 2 April (the page does not give the year).
Press reports from April 2026 put ChatGPT Plus at $20. They describe a Pro tier at $100 with "5x more Codex usage" than Plus and a Pro tier at $200 with a "20x higher usage allowance". No OpenAI page confirms these figures.
GitHub Copilot ($10–$100/month)
Copilot now meters "GitHub AI Credits" instead of premium requests. "All plans include Copilot CLI and the Copilot app."
| Plan | Price | AI credits per month |
|---|---|---|
| Free | $0 | An allowance (no number given) |
| Pro | $10 | 1,500 (1,000 base + 500 flex) |
| Pro+ | $39 | 7,000 (3,900 base + 3,100 flex) |
| Max | $100 | 20,000 (10,000 base + 10,000 flex) |
| Business | $19 per seat | 1,900 per user, pooled |
| Enterprise | $39 per seat | 3,900 per user, pooled |
On Business and Enterprise, "usage beyond the pool is charged at $0.01 USD per AI credit." A "Copilot access" option includes no credits. It costs $0 with a GitHub Enterprise Cloud licence, or $10 per user standalone.
Gemini CLI (free to start)
Gemini CLI's free tier, through a Google account, gives "1000 maximum model requests / user / day". Per-minute limits also apply, and the CLI picks the model from across the Gemini family. Google AI Pro raises that to 1,500 requests a day and Google AI Ultra to 2,000. An unpaid Gemini API key gives 250 requests a day on the Flash model only.
The same page states: "Gemini CLI was replaced by Antigravity CLI on June 18th, 2026", for the unpaid tier and Google One users. Check which CLI your plan covers before you set up a team on it.
What runs out, and what happens next
| Tool | What is metered | Past the limit |
|---|---|---|
| Claude Code (subscription) | Five-hour window plus a weekly limit, shared with Claude chat | Usage credits at standard API rates, with a monthly spending cap |
| Claude Code (API key) | Tokens at list price | Only the limits you set, such as a workspace spend limit |
| Cursor | Included usage (multiples of Plus); weekly pool for Cursor models | Prepaid credit at per-token rates; Auto Top-up and a monthly additional-usage limit |
| OpenAI Codex | Five-hour window shared by local messages and cloud tasks; weekly limits may apply | Credits (Plus, Pro), or an API key at standard API rates |
| GitHub Copilot | AI credits per month | $0.01 per AI credit beyond the pool on Business and Enterprise |
| Gemini CLI | Model requests per user per day | Paid API key: "Varies by pricing tier and model/token usage" |
Agent tools use far more context than a single chat turn. Anthropic's costs page explains why: "Claude Code sends your full conversation with every request, and each time Claude uses tools it sends another request carrying that batch of tool results."
Where the tokens in an agent session go
Anthropic's costs page shows a sample /usage line: "claude-sonnet-4-6: 1.2k input, 5.3k output, 940.0k cache read, 50.0k cache write ($0.55)". Pricing it line by line reproduces the $0.55 at Sonnet 4.6 rates. At the same token counts, Sonnet 5.5 rates halve it:
| Token type | Tokens | Sonnet 4.6 rate | Cost | Sonnet 5.5 rate | Cost |
|---|---|---|---|---|---|
| Input | 1,200 | $3.00 | $0.0036 | $2.00 | $0.0024 |
| Output | 5,300 | $15.00 | $0.0795 | $10.00 | $0.0530 |
| Cache read | 940,000 | $0.30 | $0.2820 | $0.10 | $0.0940 |
| Cache write (5-minute) | 50,000 | $3.75 | $0.1875 | $2.50 | $0.1250 |
| Total | 996,500 | $0.5526 | $0.2744 |
Rates are USD per million tokens. The token counts will not stay the same, though. Anthropic's pricing page says Claude 4.7 and later models use a tokenizer that produces "approximately 30% more tokens for the same text". Sonnet 4.6 uses the previous tokenizer. Add 30% to every line and the Sonnet 5.5 session costs about $0.36 ($0.2744 × 1.3), roughly 65% of the Sonnet 4.6 figure.
Cache reads are 94% of the tokens and the largest single cost line at Sonnet 4.6 rates. Compare tools that bill third-party models at API prices on the cache-read price as well as the headline input price. The October 2026 pricing snapshot lists cached prices per vendor.
Decision framework
If you work in the editor all day and want AI inside it, look at Cursor or Copilot. The cheapest entry points are Cursor Plus at $8 (Cursor's own models only) and Copilot Pro at $10 (1,500 AI credits). Cursor Premium at $30 is the first Cursor tier with Claude, GPT and Gemini models.
If you want to hand whole tasks to an agent, look at Claude Code (Pro $20 to Ultra $200) or Codex (inside ChatGPT plans). Both meter a five-hour window, except Claude Ultra, which has "No 5-hour limit". Both let you keep going at API rates past the plan limit.
If you want to try before paying, Gemini CLI gives 1,000 requests a day free. Copilot Free, Cursor Free and ChatGPT Free (which includes Codex) also exist, but their pages do not put a number on the allowance.
If your team is on GitHub, Copilot Business at $19 per seat comes with 1,900 credits per user, pooled across the organisation. Copilot Enterprise at $39 per seat gives 3,900 credits per user.
If you need team spend control, Claude Team and seat-based Enterprise owners can cap usage credits for the whole organisation and per member. Seat-based Enterprise adds caps per seat group. On the API, Claude Code's Console workspace "is the only workspace that supports per-user monthly spend limits." Cursor Enterprise offers pooled usage with invoicing and spend controls. Copilot pools credits and bills overage at $0.01 per credit.
If you run agents every working day, Anthropic's API average of about $13 per active day comes to $260 over 20 days. That is more than Max ($100) or Ultra ($200), provided the plan's limits hold for your workload. See API calls vs subscription for the general break-even logic.
The BYOK question
Bring-your-own-key means the tool covers the product and your API account pays for the model. The rules differ by tool:
| Tool | Own API key | Notes |
|---|---|---|
| Cursor | Not on Plus. On Teams and Enterprise, BYOK usage carries a fee | "Cursor Token Rate" of $0.25 per million tokens on eligible third-party models, which "applies on top of model API pricing for additional usage and BYOK usage". "Individual plans don't pay it." Cursor models and some others (including Claude Haiku 5.5 and GPT-5.6 Luna) are excluded |
| Claude Code | Yes: a Claude Console API key, or Amazon Bedrock, Google Cloud or Microsoft Foundry | Billed per token to your organisation. A "Claude Code" workspace is created the first time you sign in with a Console account, for cost tracking |
| OpenAI Codex | Yes, for extra local tasks | "at standard API rates" |
BYOK pays off when your per-token price is lower than what the subscription charges you, or when you already have contracted rates. Claude Code's modelPricing setting makes /usage report at contracted rates. Anthropic notes it "changes what Claude Code reports, not what Anthropic charges." Run BYOK for one month with usage logging before moving a team to it. The AI model map sorts current models by price and size, which helps when choosing what to point a key at.
What to do next
- Write down what runs out first for you today: daily requests, the five-hour window or monthly credits.
- Pick the plan whose limit matches that, using the overview table.
- Turn on a spending cap for anything billed past the plan (usage credits, prepaid credit or overage).
- After a month, compare your logged usage at API prices with the plan price.