Claude Code costs $20/month on the Pro plan ($17 annual), $100/month on Max 5x, $200/month on Max 20x, $25/seat/month on Team Standard ($20 annual), $125/seat/month on Team Premium ($100 annual). Enterprise is $20/seat plus API usage. The Free Claude tier does not include Claude Code. API access bills per token at $1/$5 per million on Haiku 4.5, $2/$10 on Sonnet 5 until August 31, and $5/$25 on Claude Opus 5. The Claude for Open Source program offers six months of Max 20x free for qualifying maintainers, a $1,200 value. Anthropic’s Claude Sonnet 5 runs at $2/$10 per million tokens until August 31, 2026, then $3/$15, and it is what Claude Code defaults to on Pro.
That short answer is also the long story of Claude Code pricing in 2026. Anthropic shipped three policy changes in 60 days and briefly tried to remove Claude Code from the Pro plan in April. It then released Claude Opus 5 on July 24, at the same $5/$25 as the model it replaced. This guide walks through every plan, every API rate, the real numbers from Anthropic’s own enterprise data, and how to keep your bill from blowing up. You will know which option fits you by the end.
The Key Takeaways
- Claude Code Pro is $20/month ($17/month annual), Max is $100 or $200/month, and Team Premium is $100/seat/month annually for full Claude Code access
- The Free Claude plan does not include Claude Code; you need Pro, Max, Team, Enterprise, or API credits
- API rates start at $1/$5 per million tokens for Haiku 4.5, sit at $5/$25 for Claude Opus 5, and top out at $10/$50 for Fable 5
- Average enterprise spend is $13 per developer per active day, or $150 to $250 per developer per month
- Prompt caching drops input costs to 10% of standard rates, and the Batch API cuts both input and output by 50%
Claude Code plans 2026: Free, Pro at $20, Max at $100 or $200, Team at $25 to $125, Enterprise
Claude Code is billed three ways. You can pay via a Claude subscription (Pro, Max, Team, or Enterprise), pay per token via the API, or use the Claude for Open Source program for free if you qualify. Most developers start on Pro at $20/month and move up only if they hit the limits.
Here is the full plan landscape as of July 2026.
| Plan | Price | Models | Claude Code limits | Best for |
|---|---|---|---|---|
| Free | $0 | Chat only | Not included | Trying Claude for chat |
| Pro | $20/mo ($17 annual) | Defaults to Sonnet 5, Opus 5 selectable | ~44K tokens / 5h window | Solo devs, occasional Claude Code use |
| Max 5x | $100/mo | Defaults to Opus 5, priority access | 5x Pro capacity | Daily Claude Code users, full-time freelancers |
| Max 20x | $200/mo | Same as Max 5x | 20x Pro capacity | Heavy agentic workflows, no rate-limit ceiling |
| Team Standard | $25/seat/mo ($20 annual) | Defaults to Sonnet 5 | Included, less usage than Premium | Teams using mostly Claude chat |
| Team Premium | $125/seat/mo ($100 annual) | Defaults to Opus 5, all models | Full Claude Code parity to Max 5x | Developer teams using Claude Code daily |
| Enterprise | $20/seat + API usage | All, with SCIM, audit logs, HIPAA | Custom | Large orgs with compliance needs |
| API (pay-per-token) | $1/$5 to $10/$50 per MTok | All Claude models | Separate billing | Burst usage, bots, CI prototypes |
Annual Pro is billed as $200 up front, which Anthropic advertises as $17/month. That is $40 less than twelve monthly payments of $20. Team Premium annual at $100/seat is the same price as Max 5x, with the bonus of admin controls, SSO, and shared usage pools.
Was Claude Code removed from the Pro plan? The April 2026 confusion
Short answer, no, Claude Code is back on Pro and that is the official policy on the live pricing page right now. But for several days in late April 2026, Anthropic ran what its Head of Growth called a “small test on ~2% of new prosumer signups” that gated Claude Code behind Max only. Screenshots leaked, Reddit and Hacker News exploded, and the change was reverted within hours, as documented in Simon Willison’s blog post.
The takeaway is twofold. First, your $20 Pro subscription still includes Claude Code as of July 2026. Second, Anthropic has tested gating it before, so the price you see today is not guaranteed forever. Lock in annual billing if Pro is essential to your workflow.
Which Claude model your plan actually runs
This is the part most pricing guides miss, and it changes what you are buying. Claude Code does not run one default model for everyone. Per Anthropic’s model configuration docs, the default setting resolves differently depending on how you pay.
Max, Team Premium, Enterprise pay-as-you-go and direct Anthropic API accounts default to Claude Opus 5. Pro, Team Standard and Enterprise subscription seats default to Claude Sonnet 5. So a $20 Pro user following a $100 Max user’s tutorial is running a different model, and the same prompt can produce different results and different costs.
You can still switch. Run /model to pick Opus 5 on Pro, or /model fable for Fable 5, which is not the default on any account type. There is also an opusplan mode that plans with Opus and executes with Sonnet, which is the cheapest way to get Opus-quality architecture decisions without paying Opus rates for every edit.
Claude Code plan breakdown
Each tier has a different ceiling and a different audience. The right pick is almost always either Pro or Max 5x, with API as a fallback for spiky workloads.
Free at $0 (no Claude Code)
The Free Claude plan at $0 does not include Claude Code. Anthropic’s pricing FAQ puts it plainly, “Claude Code is included in all paid plans”, so the CLI needs Pro, Max, Team, Enterprise or API credits. The closest free path is the Claude for Open Source program for maintainers, or API starter credits for new accounts. Our guide on whether Claude Code is free breaks down each free path in full.
Pro at $20/month
Pro at $20/month ($17/month annual) is where most solo developers live. It includes Claude Code in the terminal, web, and desktop, defaults to Sonnet 5 with Opus 5 selectable, and gives you a token budget that handles a few focused coding sessions a day. Anthropic does not publish exact Pro token numbers publicly, but community instrumentation puts it at roughly 44,000 tokens per 5-hour window, enough for 10 to 40 prompts depending on context size.
Pro’s catch is the rate limit. If you run agentic loops or open three Claude Code sessions at once, you will hit the wall before lunch. That is when Max stops being optional.
Max 5x at $100/month
Max 5x at $100/month gives you five times the Pro usage, priority access during peak demand, and Claude Opus 5 as the default model rather than Sonnet 5. It is the sweet spot for full-time developers who use Claude Code as their default coding partner. Anthropic’s own data says 90% of Claude Code users stay under $30 per active day, which Max 5x covers comfortably.
Max 20x at $200/month
Max 20x at $200/month is the no-rate-limit tier. Anthropic’s own pricing page lists Max as “from $100 per month” and describes the two tiers as 5x and 20x more usage than Pro. You get 20 times Pro capacity, which one developer in a public Reddit teardown calculated as roughly $5,000 of equivalent API compute if you actually use it. If you are running multiple Claude Code instances, agent teams, or all-day pair programming, this is cheaper than the API.
Team Standard at $25/seat/month
Team Standard at $25/seat monthly ($20/seat annually) is built around Claude chat, and it does include Claude Code with less usage than Premium. Anthropic sells Team “for teams of 2 to 150”, so the real floor is 2 seats, or $40/month annual ($50 monthly). If your team is mostly writing and analysis with one or two coders, Team Standard plus a single Max seat is often cheaper than upgrading everyone to Premium.
Team Premium at $125/seat/month
Team Premium at $125/seat monthly ($100/seat annually) is the developer tier with full Claude Code parity to Max 5x, plus admin controls, SSO, and shared usage pools. It is also the seat type that defaults to Opus 5. With the same 2-seat floor, the entry cost is $200/month annual ($250 monthly), and Anthropic lets you mix seat types inside one Team plan. At equivalent annual pricing to Max 5x, Team Premium adds the admin layer for free.
Enterprise at $20/seat plus API usage
Enterprise is $20 per seat per month plus usage billed at API rates, in Anthropic’s own words, with SCIM provisioning, audit logs, HIPAA-readiness, and a compliance API. There is a self-serve version you can start today and a sales-assisted one for custom terms. There is no public price ceiling because enterprise contracts are negotiated. If your org needs custom rate limits, dedicated support, or volume discounts, contact Anthropic sales rather than stacking Max seats.
Claude Code API pricing per token
If you are running Claude Code in CI, in a bot, or in workloads that spike unpredictably, the API is often cheaper than a fixed subscription. API access bills purely on tokens, with discounts for prompt caching and batch processing.
| Model | Input ($/MTok) | Output ($/MTok) | Cache Read ($/MTok) | Notes |
|---|---|---|---|---|
| Claude Fable 5 | $10 | $50 | $1 | Most capable, not the default anywhere; select it with /model fable |
| Claude Opus 5 | $5 | $25 | $0.50 | Current flagship, default on Max and Team Premium; 1M context, 128k max output |
| Claude Opus 4.8 | $5 | $25 | $0.50 | Previous flagship, identical rates |
| Claude Opus 4.7 | $5 | $25 | $0.50 | Older flagship, identical rates |
| Claude Sonnet 5 | $2 | $10 | $0.20 | Default on Pro and Team Standard; introductory pricing through Aug 31 2026, then $3/$15 |
| Claude Sonnet 4.6 | $3 | $15 | $0.30 | Previous Sonnet, leaner older tokenizer |
| Claude Haiku 4.5 | $1 | $5 | $0.10 | Cheapest, good for simple subagent tasks |
MTok = million tokens. Output is always 5x more expensive than input on every Claude model, which is why Claude Code’s auto-compaction matters so much. Cache reads are 10% of the base input price across the board.
The hidden gotcha is the newer tokenizer. Per Anthropic’s pricing docs, Claude 4.7 and later models, which now includes Opus 5 and Sonnet 5, use roughly 30% more tokens for the same text than older models. Your sticker price per token is unchanged, but your real bill rises about 30% on tasks routed to those models if you do not adjust. Sonnet 4.6 and earlier keep the older, leaner tokenizer.
Two discounts can drop your bill dramatically. Prompt caching lets you reuse system prompts and large contexts at 10% of the input price, so cached reads pay off after a single hit on the 5-minute cache. The Batch API cuts both input and output by 50%, suitable for non-time-sensitive workloads like overnight refactors. Both stack with each other.
Real-world Claude Code costs
Anthropic publishes real numbers in its cost-management documentation. Across enterprise deployments, average Claude Code spend is $13 per developer per active day, or $150 to $250 per developer per month, with 90% of users staying under $30 per active day. Those are full-time professional figures; solo developers spend less.
Here is what real spend looks like by user type.
| User type | Daily token use | Daily $ on API | Monthly $ on API | Recommended plan |
|---|---|---|---|---|
| Light (occasional bug fixes, side projects) | 200K-500K | $2 to $5 | $50 to $100 | Pro $20 |
| Medium (daily features, multi-file edits) | 1M-2M | $6 to $12 | $130 to $260 | Max 5x $100 |
| Heavy (agentic loops, multi-instance) | 3M-8M+ | $20 to $60+ | $400 to $1,200+ | Max 20x $200 |
| Team (5 devs, mixed usage) | 10M-25M | $50 to $150 | $1,000 to $3,000 | Team Premium $500+ |
Subscriptions almost always win on Claude Code at professional volume. One developer who instrumented their network usage and shared it on Reddit reported 10 billion tokens over 8 months. Equivalent API cost would have been over $15,000; their actual Max bill was around $800, a 93% saving.
The math reverses for spiky usage. If you only run Claude Code two days a week, the API plus a free Claude account will likely beat any subscription.
How to get Claude Code free or cheap
There are three paths to Claude Code without paying full sticker.
The Claude for Open Source program is the biggest one. Qualifying open-source contributors get six months of Max 20x free, a $1,200 value, and you apply via Anthropic’s Open Source page. Anthropic rewrote the eligibility rules, and the current version is far broader than the star-count thresholds that circulated earlier. There are now five separate routes in, and you only need one.
The first route is maintainers and library authors, defined as 500 or more dependent repos, 100 or more dependent packages, or 200,000 or more combined monthly downloads across any registry. The second is core contributors. That means a listed committer or maintainer on a recognised foundation or language project, such as CPython, the Rust team, the Node.js TSC, an Apache PMC, CNCF, Kubernetes, the Linux kernel, Django or Rails.
The other three routes reward activity rather than scale. Active contributors qualify with 100 or more pull requests merged into repos they do not own in the last 12 months. Community builders qualify if one of their repos has had 20 or more unique external contributors with merged pull requests in 12 months. Critical infrastructure qualifies at an OpenSSF criticality score of 0.4 or above. Anthropic also invites people who fit none of these to apply anyway and explain what they maintain.
When the six months end, the free subscription simply stops. If you were paying before, your old plan and rate resume unless you cancel; otherwise the account drops to the free plan.
New API users get free starter credits (reported at around $5, no credit card required), enough to run Claude Code for a few hours of light coding. It is the cheapest way to actually try the CLI without committing to a $20/month subscription. After the credits are spent, you decide whether to top up the API balance or move to Pro.
Local alternatives are the third path. Tools like OpenCode and Ollama-routed proxies let you point a Claude Code-style CLI at locally-run models like Qwen2.5-Coder or DeepSeek-Coder. Speed and quality are below the real Claude Code, but cost is $0 plus electricity. If you are a hobbyist or you have GPU hardware sitting idle, this is the genuine free path. For a hosted frontier model with the same Anthropic-API drop-in compatibility, see our Qwen3.7-Max review. For a ranked rundown of the strongest no-cost options, see our guide to the best free AI for coding.
Subscription vs API: which saves you money?
The break-even between a subscription and the API depends on how many hours per day you actually use Claude Code. Here is a clean rule of thumb based on the Anthropic per-developer-day numbers.
If you use Claude Code less than one focused hour per day, the API plus a free Claude account is cheapest. You will burn maybe $1 to $3 per day, which is under the $20 Pro monthly fee.
If you use it one to four hours per day, Pro at $20/month is the right call. Anthropic’s data says median active users spend $6 to $13 per active day on equivalent API, so Pro pays for itself in two to four days.
If you use it all day, every workday, Max 5x at $100/month matches enterprise averages. You stop watching the meter, and the per-day Anthropic average ($13) times 20 workdays equals $260 of API equivalent for a $100 fee.
If you run multiple instances or agent teams, Max 20x at $200/month is the only sane choice. Solo developers rarely need it. Teams almost always do. Long-running agents also gain from Dreaming for Claude Code agents, which carries learned context between runs.
How to cut your Claude Code bill
Even on Max 20x, there are cost-control levers worth pulling. Anthropic’s docs say teams that apply the optimization playbook see 40% to 85% reductions in token usage, and several of the tactics also speed Claude Code up.
The biggest win is prompt caching. Cached reads cost 10% of input price, so any system prompt or document you reuse should be cached. Claude Code does this automatically for CLAUDE.md and recently-read files, but you can extend it by keeping conversation history short and stable.
The second-biggest win is picking the right model. Sonnet 5 handles the vast majority of coding tasks at $2/$10 per million tokens (introductory, rising to $3/$15 after Aug 31 2026), versus Opus 5’s $5/$25 and Fable 5’s $10/$50. Use Opus only for hard architectural decisions, multi-step reasoning, or critical refactors. Run /model to switch mid-session, and configure subagents to use Haiku 4.5 for simple lookups.
The third lever is proactive context management. Run /clear when you switch tasks instead of letting stale context pile up. Use /compact to summarize long sessions. Run /usage to check token spend live. The /cost command shows API session cost as a local estimate; on Max or Pro it is informational, since you are not billed per call.
Other cost levers worth knowing. The Batch API halves token costs on non-urgent work like overnight refactors. Plan mode (Shift+Tab) makes Claude propose an approach before writing code, preventing expensive wrong directions. Hooks can pre-filter verbose output (a 10,000-line log becomes 50 lines of grep matches before Claude ever sees it). Skills load on demand instead of bloating your CLAUDE.md.
For teams, set workspace spend limits in the Claude Console. Several large enterprises run LiteLLM as a separate cost tracker on top of Bedrock or Vertex deployments. Both are documented in Anthropic’s cost docs.
Claude Code vs Cursor vs GitHub Copilot pricing
Claude Code is not the only paid coding agent. Here is how it compares to the two biggest competitors plus OpenAI’s Codex. Google’s agent-first Antigravity platform is a newer rival that is free during its public preview, worth watching against Claude Code’s paid tiers.
| Tool | Entry plan | Top individual plan | Free tier | Flagship model |
|---|---|---|---|---|
| Claude Code | $20/mo (Pro) | $200/mo (Max 20x) | API starter credits, OSS program | Opus 5, Sonnet 5, Fable 5 |
| Cursor | $20/mo (Individual) | Ultra, see our Cursor guide | Hobby tier (limited) | Bring-your-own or Cursor’s own |
| GitHub Copilot | $10/mo (Pro) | $100/mo (Max) | Free plan, 2,000 completions | Opus 5, GPT-5.6 Sol, Gemini 3.6 Flash |
| OpenAI Codex | $20/mo (ChatGPT Plus) | ChatGPT Pro | API credits only | GPT-5 series |
Claude Code is mid-priced at the entry tier but highest at the top end because Max 20x carries serious capacity. GitHub Copilot is still the budget pick at $10/month, and its free plan is permanent rather than a trial, giving you 2,000 completions a month. Copilot has also added a $100/month Max tier for sustained agent work, and its Pro plan now runs Claude Code and Codex as third-party agents inside Copilot. Cursor’s per-token pricing lets you use Claude or GPT models inside Cursor’s editor at API rates plus a small markup; our full Cursor pricing 2026 guide breaks down every tier. OpenAI Codex on Mac added desktop computer use in April 2026, putting it in direct competition with Claude Code’s terminal-first workflow.
One caveat on the table. OpenAI’s public pricing pages did not load for us while this guide was being checked, so we are not printing a ChatGPT Pro figure rather than repeating one we could not verify. The honest verdict is Claude Code wins on agentic depth, Cursor wins on IDE polish (students can claim 1 year free), Copilot wins on price, and Codex wins if you live inside ChatGPT.
Should you pay for Claude Code?
For solo developers writing daily, Pro at $20/month is the right starting point. You will pay back the subscription in your first focused coding session. If you find yourself hitting limits twice a week or more, jump to Max 5x at $100/month without overthinking it; the payback math holds at moderate professional volume.
For teams of two or more developers, Team Premium annual at $100/seat beats stacking individual Max plans. You get admin controls, SSO, shared usage, and Opus 5 in front of every seat by default. For everyone else, the API with a starter credit is the cheapest first step.
If you want Claude for chat, files, and writing on your Mac rather than coding, you do not need Claude Code at all. Fello AI gives you Claude alongside ChatGPT, Gemini, Grok, and DeepSeek in a single Mac-native app for $9.99/month, which is half the price of Pro and covers five model families instead of one. For the pure chat experience compare it against the full Claude pricing guide and how Claude on the Mac stacks up against ChatGPT and Gemini before committing.
For everything else, including pricing comparisons across the full AI tool landscape, our complete AI pricing comparison for 2026 covers every major model alongside Claude. If you want to use images with Claude Code specifically, our guide on Claude Code images walks through the supported workflows on every OS.
FAQ
Is Claude Code free?
No. Claude Code is not on the free Claude plan. The cheapest paid path is Pro at $20/month, and the cheapest free path is the Claude for Open Source program, which gives qualifying maintainers six months of Max 20x free. New API users also get free starter credits (reported at around $5).
How much does Claude Code cost per month?
Pro is $20/month, or $200 billed annually. Max is $100 or $200/month, and Team Premium is $100 or $125/seat/month. The API has no monthly minimum and bills per token, from $1/$5 on Haiku 4.5 to $5/$25 on Opus 5. Average enterprise usage runs $150 to $250/developer/month per Anthropic’s own data.
Is Claude Code worth $100 a month?
For full-time developers who hit Pro rate limits regularly, yes. Max 5x at $100/month gives you roughly 5x the Pro usage plus priority access during peak demand. The break-even versus equivalent API cost is around two to three hours of focused use per day.
How is Claude Code billed?
Through your existing Claude subscription (Pro, Max, Team, or Enterprise) for fixed monthly billing, or through the Anthropic API for pay-per-token billing. Subscriptions count Claude Code usage against the same shared token budget as Claude chat.
Which Claude model does Claude Code use by default?
It depends on your account. Max, Team Premium, Enterprise pay-as-you-go and Anthropic API accounts default to Claude Opus 5, while Pro, Team Standard and Enterprise subscription seats default to Claude Sonnet 5. Run /model to switch, and note that Fable 5 is not the default on any plan.
What is the cheapest way to use Claude Code?
The Claude for Open Source program at $0 for six months if you qualify. After that, free API starter credits (reported at around $5) for new accounts, then Pro at $200 billed annually, which works out at $17/month. Local alternatives like Ollama-routed CLIs cost nothing but trade quality and speed for the savings.
