ClaudeHow · facts verified 30 Sept 2026
What should I use, and what will it cost? — practical answers for building on Claude.
Every calculator computes from your real usage, not a vendor's example. Every fact renders from a dated, sourced module — never typed by hand into a page.
The lineup right now
The newest model in each line, per million tokens. Every other model — and batch, cache-write and fast-mode rates — is on the full pricing table.
| Model | Input | Output | Cache read | Context | Effort default |
|---|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 | 1M | high |
| Opus 5.5 | $4 | $20 | $0.20 | 1M | medium |
| Sonnet 5.5 | $2 | $10 | $0.20 | 1M | high |
| Haiku 4.5 | $1 | $5 | $0.10 | 200K | — |
What changed lately
Dated, sourced, and linked to the page that works out what it means for you.
28 Sept 2026 · New model
22 Sept 2026 · New model
22 Sept 2026 · Plans
16 Sept 2026 · Plans
1 Sept 2026 · Pricing
Calculators that start from your numbers
Subscription vs API cost calculator
Model your actual Claude usage both ways — flat subscription and metered API — and see the break-even point, not a vendor's marketing comparison.
Token & cost estimator
Paste text to see its real Claude token count and cost across every current model — and why a tiktoken-based estimate undercounts it.
Prompt-caching savings calculator
Enter your prefix size and request pattern to see whether the 5-minute or 1-hour cache TTL wins, and what it saves per month.
Model picker
Task shape, budget ceiling, latency tolerance and context size in — a recommended model and effort level out, with the reasoning shown.
Context-window planner
See how much of a 1M-token window your codebase or document set consumes, what's left for output, and where compaction starts paying for itself.
Usage tracker
Log your own usage or paste API usage blocks to see burn rate, projected month-end spend and a cache-hit-ratio read. Nothing leaves your browser.
Batch vs realtime calculator
The Batch API runs at half the standard price with a turnaround measured in hours. Given your volume and deadline, is batching worth it?
Effort & thinking cost estimator
Effort runs low through max and is the main lever on both spend and quality. See the cost curve and where each level is the right call.
Reference readers come back for
Claude API pricing: every model, every rate
Input, output, cache-write, cache-read, batch and fast-mode prices for every current Claude model, plus the modifiers that change the bill.
Claude plans and prices: Pro, Max, Team and Enterprise compared
Every Claude plan's US list price, what it includes, how usage compares, who gets the Fable models, and where plan usage turns into API-rate billing.
Claude model retirement dates
Which Claude models are active, deprecated or retired, the earliest date each active model can retire, and where retired ids still answer.
What changed in Claude: a dated log for developers
Every model launch, price change, retirement and plan change that moves a developer's bill or breaks their code — dated, sourced, newest first.
What should I use, and what will it cost?
Anthropic's own documentation is thorough about what the API does. It is deliberately silent on the question every working developer actually has: given how I use Claude, should I be on a flat subscription or metered API, which model, and what will next month's bill look like? No vendor publishes that comparison, because the honest answer is "it depends on your usage" and a vendor has no incentive to build you a calculator that might recommend a cheaper plan.
ClaudeHow exists to answer that question with your numbers, not a marketing example. Every tool here takes your actual usage as input — messages per day, typical prompt and output length, how much context you re-send, your request pattern — and computes an answer instead of asserting one.
Why the numbers can be trusted
Every price, context window, cache minimum and effort level published on this site comes from a single facts module with a recorded verification date and source, never from prose typed by hand. A script fails this site's own build if any figure goes stale, so a page here cannot quietly drift out of date the way a blog post or a cached search result can — and where the underlying source doesn't fully answer a question, this site says so on the page rather than filling the gap with a guess. See the about page for the full detail behind that discipline.
Plans and prices, side by side
Claude plans explained lists Anthropic's US list price for every plan alongside what each one includes — relative usage, Claude Code, which plans reach the top Fable models, and where a plan's usage starts billing at API rates. The subscription-vs-API calculator still starts with an empty price box, because what you pay depends on currency, tax and billing period; tap a list price to start from it, or enter your own.
What you will not find here
A restatement of Anthropic's own docs. Every page here computes something, compares something measurable, or catalogues a specific mistake with the one-line fix, because a page that repeats the documentation earns nothing a reader couldn't get faster from the source.
Start with the calculator that started this site
The subscription vs API cost calculator models your actual workload both ways and names the break-even point in plain language. If you only use one tool on this site, that is the one worth your time.