Reasoning and writing
Claude pricing,
plans and models.
Careful long-form writing, reasoning through complex material, and disciplined coding or document work.
Claude is a general AI assistant. Careful long-form writing, reasoning through complex material, and disciplined coding or document work. This guide covers the whole path in one place: official access, a first session that produces something reviewable, the checks that make output trustworthy, and the permissions worth limiting before you connect real work.
AI should make the work easier to inspect. If the workflow removes the source, the owner, or the review step, redesign the workflow.
Access & setup
Find, install, and sign in to Claude
Get into the official Claude experience with the right account and a setup you understand.
- 01
Start at https://claude.ai/ and confirm the domain before entering account or payment information.
- 02
Availability: macOS and Windows.
- 03
Anthropic documents macOS 11 or later and Windows 10 or later for Claude Desktop.
- 04
Sign in with the account you intend to keep using, then review plan, data, notification, and permission settings.
- 05
Run one low-risk test task before connecting sensitive files, repositories, or workspace data.
- Use official download pages.
- Review permissions during setup.
- Keep installers and applications updated.
First session
Your first useful Claude session
Learn the interaction loop using a small task with a clear outcome.
- 01
Install Claude Desktop or sign in at claude.ai with the account you intend to use.
- 02
Give Claude a clear outcome first, followed by only the context that changes the answer.
- 03
Specify the response format and ask it to restate the task if the work is complex.
- 04
For recurring desktop work, review Extensions in Settings and connect only the tools or data you actually need.
- State the outcome before the background.
- Provide the real source material.
- Review the result before expanding the task.
Quality control
Check the quality of Claude output
Turn a plausible draft into work you are willing to stand behind.
- 01
Restate the original outcome and acceptance criteria before reviewing the response.
- 02
Mark every factual claim, calculation, quotation, or figure that could materially affect the result.
- 03
Check those items against the original source, a primary reference, or direct inspection.
- 04
Ask the tool to list its assumptions and uncertainty, but never rely on self-review as the only check.
- 05
Complete a human edit for relevance, tone, privacy, accessibility, and the exact delivery format.
- Verify consequential claims independently.
- Fluency is not accuracy; confident phrasing is not evidence.
- Keep a record of sources and review decisions.
Privacy & permissions
Use Claude safely
Use the platform productively without giving it unnecessary access or authority.
- 01
Classify the task and its data before uploading or connecting anything.
- 02
Review the product, account, organisation, retention, and training settings that apply to you.
- 03
Grant the narrowest file, workspace, microphone, screen, or integration permissions needed.
- 04
Remove secrets and personal or regulated information unless your approved policy explicitly permits it.
- 05
Review the output and revoke permissions or disconnect sources that are no longer needed.
- Desktop extensions can access connected tools and data. Use the smallest permission scope that accomplishes the task, and independently check claims or code before using them.
- Follow your organisation's approved-use policy.
- Never treat fluent output as authorization to act.
Models & pricing
What Claude costs, model by model.
Anthropic's line-up runs from Haiku 4.5 for high-volume work to Fable 5.1 at the frontier, with Sonnet 5 as the general-purpose default and Opus 5 for the hardest reasoning.
| Model | Role | Input | Output |
|---|---|---|---|
| Fable 5.1 | Frontier tier | $10.00 | $50.00 |
| Opus 5 | Hardest reasoning and long agentic work | $5.00 | $25.00 |
| Sonnet 5 | General-purpose default | $2.00 | $10.00 |
| Haiku 4.5 | Fast, high-volume, low cost | $1.00 | $5.00 |
Figures verified 2 September 2026 against Claude platform pricing ↗. Model pricing changes frequently — confirm before you budget.
| Provider | Frontier tier | Balanced default | Low-cost tier |
|---|---|---|---|
| OpenAI | GPT-5.6 Sol — $5 / $30 | GPT-5.6 Terra — $2 / $12 | GPT-5.6 Luna — $0.20 / $1.20 |
| Anthropic | Fable 5.1 — $10 / $50 | Sonnet 5 — $2 / $10 | Haiku 4.5 — $1 / $5 |
| Gemini 3.1 Pro — $2 / $12 | Gemini 3.1 Pro — $2 / $12 | Gemini 3.5 Flash-Lite — $0.15 / $1.25 |
Cross-provider comparison verified 2 September 2026. Sources: OpenAI API pricing ↗ · Claude platform pricing ↗ · Gemini API pricing ↗
Common questions
Claude questions people actually ask.
How much does the Claude API cost per million tokens?
As of September 2026, from $1 input / $5 output per million tokens on Haiku 4.5 up to $10 / $50 on Fable 5.1. Sonnet 5 sits at $2 / $10 and Opus 5 at $5 / $25. Batch processing halves these rates and prompt caching cuts cached input by 90%.
Did Claude Sonnet 5 get more expensive?
The sticker price did not rise — a scheduled increase to $3 / $15 on 1 September 2026 was cancelled, leaving Sonnet 5 at $2 / $10. However, Sonnet 5 uses a new tokenizer that produces roughly 1.0 to 1.35 times more tokens than Sonnet 4.6 for identical text, so your real bill per document can rise even at an unchanged rate. Measure tokens, not pages.
Which Claude model should I use?
Sonnet 5 is the intended default and handles most production work at $2 / $10. Use Haiku 4.5 for classification, extraction, and other high-volume, low-judgement tasks. Reserve Opus 5 and Fable 5.1 for work where a measurable quality gap justifies five to ten times the cost.
Core workflows
Step-by-step ways to use Claude for the work it does best.
Each workflow below has its own outcome, steps and review checkpoints.
Use Claude to work through substantial documents while preserving the argument, evidence, and uncertainty.
↓ 02 WorkflowDraft and revise long-form writing with ClaudeBuild an editorial workflow that protects your voice instead of producing generic AI prose.
↓Workflow 01
Turn long source material into a clear brief
Use Claude to work through substantial documents while preserving the argument, evidence, and uncertainty.
Outcome: Create a concise brief that can be traced back to the source.
- 01
Provide the documents with a sentence explaining their origin, audience, and relationship to one another.
- 02
Ask Claude to map the documents first: sections, claims, evidence, contradictions, and missing context.
- 03
Define the brief's reader, length, decision, and required headings.
- 04
Request source references for each key conclusion and a separate list of inferences.
- 05
Review the decisive passages, then ask Claude to revise only the claims your review changed.
What good use looks like
- Ask for an outline before prose.
- Keep quotations short and traceable.
- Do not combine unrelated source sets in one thread.
Workflow 02
Draft and revise long-form writing with Claude
Build an editorial workflow that protects your voice instead of producing generic AI prose.
Outcome: Develop a strong draft through deliberate stages.
- 01
Give Claude a representative sample of your writing and identify the qualities to preserve.
- 02
Describe the reader, purpose, argument, evidence, length, and phrases or habits to avoid.
- 03
Ask for three different outlines and choose one before drafting.
- 04
Revise in passes: argument first, structure second, sentence quality third, proofing last.
- 05
Read the final draft aloud and restore any specificity or rhythm that disappeared.
What good use looks like
- Give feedback with examples.
- Revise one dimension at a time.
- Keep final editorial responsibility with a person.
Official references
Check the current product documentation.
Features, plan limits, availability, and data controls change. These official pages are the starting points used for this guide.