Tooling
Claude Pricing in 2026: Every Model, Plan, and Cost Lever Explained
Every Claude model rate, every subscription tier, and the caching and batch levers that decide your real cost.
If you are budgeting for Claude in 2026, there are two separate pricing worlds: the per-token API and the flat-rate subscriptions. This guide covers both, plus the cost levers that matter more than which model you pick.
Last checked: September 8, 2026, against claude.com/pricing. Prices move often, so confirm before you commit a budget.
Claude API pricing by model
Anthropic runs a four-tier lineup, from the Mythos-class flagship down to the cheap, fast workhorse. Rates are per million tokens.
| Model | Input ($/M) | Output ($/M) | Best for |
|---|---|---|---|
| Claude Fable 5.1 | $10 | $50 | Hardest agentic, coding, research |
| Claude Opus 5 | $5 | $25 | Flagship reasoning for most teams |
| Claude Sonnet 5 | $2 | $10 | Balanced daily driver, newly cheaper |
| Claude Haiku 4.5 | $1 | $5 | High-volume, fast, cheap tasks |
The headline of this refresh is the mid-tier: Sonnet 5 replaced Sonnet 4.6 and cut the price from $3/$15 to $2/$10 at the same time. If your workload ran on Sonnet 4.6, switching cuts your bill by a third before any prompt optimization.
Fable 5.1 is the public, safeguarded deployment of Anthropic’s Mythos-class weights, and at $10/$50 it now sits at exactly the same price as OpenAI’s GPT-6 Astra, which makes cross-provider flagship comparisons unusually clean this cycle. Opus 5 replaced Opus 4.8 at the same $5/$25 and remains the default for hard reasoning. Haiku 4.5 stays the volume tier where speed and cost beat maximum reasoning.
For reference against older search results: Opus 4.8 ($5/$25) and Sonnet 4.6 ($3/$15) were the current models until this generation rolled over; their prices no longer apply to the live lineup.
The cost levers that actually move your bill
The headline rate is only the start. Three levers matter more:
Prompt caching cuts cached input by about 90%. Any workflow that resends the same context (a system prompt, a codebase, a long document) should cache it. In agents and chat, this is the single biggest saving available.
The Batch API is 50% cheaper. For asynchronous work (bulk generation, evals, pipelines), batch halves the cost.
Output is 5x input. Output is the expensive half, so tighter, structured outputs cut bills faster than switching models.
Claude subscription plans
If you use Claude through the app or Claude Code, you are on a subscription rather than token billing.
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Chat, web search, memory, file creation, skills, connectors |
| Pro | $20/mo ($17 annual) | Adds Claude Code, Cowork, Design, Science |
| Max 5x | $100/mo | ~5x Pro usage |
| Max 20x | $200/mo | ~20x Pro usage |
| Team Standard | $25/seat ($20 annual) | Collaboration, min seats apply |
| Team Premium | $125/seat ($100 annual) | Much higher usage per seat |
| Enterprise | $20/seat + usage-scaled | SCIM, audit logs, HIPAA-ready option |
Plan prices held steady since June, but they bundle noticeably more product now: Pro includes Cowork, Design, and Science alongside Claude Code.
Anthropic’s own data puts the average Claude Code user near $6/day, with 90% under $12/day. At full-time usage that is roughly $100-$200/month, which is exactly where Max sits. The rule of thumb: heavy daily users save on Max; light or spiky usage is cheaper on the API.
Which Claude plan should you pick?
If you are an individual using Claude occasionally, Pro is plenty. If you live in Claude Code all day, jump to Max 5x and move to 20x only if you keep hitting limits. Teams that need high usage per seat go Team Premium; everyone else sits on Standard. Enterprises with governance, audit, or data-residency needs go custom.
FAQ
How much does the Claude API cost in 2026? Per million tokens: Fable 5.1 $10/$50, Opus 5 $5/$25, Sonnet 5 $2/$10, Haiku 4.5 $1/$5. Caching cuts cached input ~90%; batch is 50% off.
What is the cheapest Claude model? Haiku 4.5 at $1/$5 per million tokens. Sonnet 5 at $2/$10 is the value pick for production.
Is Claude Pro or Max better value? Pro for most individuals; Max if you use Claude Code heavily, since it beats per-token billing at full-time usage.
How much is Claude for teams? Team Standard $25/seat ($20 annual), Team Premium $125/seat ($100 annual), Enterprise $20/seat plus usage.
This page is part of our full AI model pricing and comparison guide. See also OpenAI GPT pricing, Gemini pricing, and Grok pricing, or jump to the best AI model for coding and the cheapest AI model. For optimizing content for Anthropic’s models, see our Claude SEO guide.
Verify against the official Claude API pricing and Claude plans pages.
