Claude is a family of large language models built by Anthropic. As of August 2026 there are four models you can actually call — Fable 5, Opus 5, Sonnet 5 and Haiku 4.5 — reachable through the claude.ai chat apps, the desktop and mobile apps, and the Claude API. This page is the map: what each model is for, what it costs, and where to go next.
This site is an independent reference. It is not Anthropic, it does not sell subscriptions and it has no login form. Everything below was checked against Anthropic's own documentation on 21 August 2026.
#Claude at a glance, August 2026
- Made by
- Anthropic PBC
- Current models
- Fable 5, Opus 5, Sonnet 5, Haiku 4.5
- Largest context
- 1,000,000 tokens
- Cheapest paid plan
- Pro — $20/month, or $200/year
- API entry price
- $1 / $5 per million tokens (Haiku 4.5)
- Where it runs
- Web, iOS, Android, macOS, Windows, Linux, terminal, API
#What is Claude, and what is it actually good at?
Claude is a general-purpose assistant: you give it text, images or files and it writes, reasons, codes, analyses and — increasingly — acts. The last two years have shifted its centre of gravity from “chatbot that writes well” to “model that can run a long task on your behalf”, which is why Anthropic's newest release, Fable 5, is described as a model for long-running agents rather than a better conversationalist.
Where Claude has a genuine, measurable edge in August 2026:
- Agentic coding. Claude sits at the top of the independent Artificial Analysis Intelligence Index, with Opus 5 scoring 63, and holds six of the top twelve places on arena.ai's Elo leaderboard.
- Novel-problem reasoning. On ARC-AGI-3, verified independently by the ARC Prize foundation, Opus 5 scores 30.16% against GPT-5.6 Sol's 7.78% — roughly a four-fold gap.
- Long, careful documents. A 1M-token context plus 128k output tokens means whole codebases, contract sets and research corpora fit in one call.
And where it does not — stated plainly, because you should know before you pay:
- No image, video, audio or music generation at all, and no video or audio input either. If you need those, ChatGPT or Gemini are the answer. See Claude vs ChatGPT.
- Weakest long-context retrieval of the big three. On GDM-MRCR v2, Gemini scores 97.0% to Sonnet 5's 81.5% — a large window is not the same as reliable recall. See Claude vs Gemini.
- The most expensive and the slowest frontier family. Fable 5 generates around 75 tokens per second; Gemini 3.7 Flash manages roughly 3,900.
Our full assessment, including who should not buy it, is on the Claude review page.
#Which Claude model should you use?
Four models, four jobs. Most people should default to Sonnet 5 and only move up when a task genuinely fails.
| Model | API ID | Context / output | Price in / out per MTok | Use it for |
|---|---|---|---|---|
| Fable 5 | claude-fable-5 | 1M / 128k | $10 / $50 | Long-horizon agents where a failure costs more than the tokens |
| Opus 5 | claude-opus-5 | 1M / 128k | $5 / $25 | Complex agentic coding, enterprise work, newest knowledge (May 2026) |
| Sonnet 5 | claude-sonnet-5 | 1M / 128k | $2 / $10 | The default. Best balance of speed, intelligence and price |
| Haiku 4.5 | claude-haiku-4-5 | 200k / 64k | $1 / $5 | Classification, extraction, routing, anything high-volume |
Two things surprise people. First, the 1M-token context is the default on Fable 5, Opus 5 and Sonnet 5 — no beta header, no long-context price tier. Second, knowledge cutoffs are not in version order: Opus 5 knows the world to May 2026, while Fable 5 and Sonnet 5 stop at January 2026. If recency matters more than raw capability, Opus 5 is the newer brain.
The full lineup, including retired versions and platform-specific model IDs, is on the Claude models page. If you are choosing between tiers for a specific workload, the Haiku vs Sonnet vs Opus comparison works through cost-per-task rather than benchmarks.
#How much does Claude cost?
Two separate price lists: subscriptions for using the apps, and per-token rates for the API. They do not overlap — an API key does not give you Pro, and Pro does not give you API credit.
| Plan | Price | Billing | Best for |
|---|---|---|---|
| Free | $0 | — | Trying Claude; light daily use |
| Pro | $20/month or $200/year | Monthly or annual | Individuals who also want Claude Code and Cowork |
| Max 5× | $100/month | Monthly only | Heavy daily users hitting Pro limits |
| Max 20× | $200/month | Monthly only | All-day agentic coding |
| Team | From $20/seat/month | Monthly or annual | 2–150 seats with admin controls |
| Enterprise | $20/seat/month + usage at API rates | Annual only | SSO, SCIM, audit logs, custom retention |
Anthropic does not publish an absolute message count for any plan. Usage is governed by a rolling five-hour session window and a weekly limit, with Opus models tracked on a separate weekly counter, and chat, Claude Code, Cowork, Design and Excel all drawing from one shared pool. Any site quoting “about 45 messages every five hours” is repeating a figure that is not in current documentation. The mechanics are explained on the pricing page.
On the API side the levers that actually change your bill are prompt caching (cache reads cost 0.1× base input) and the Batch API (a flat 50% discount). Both are worked through with numbers on the Claude API page.
#Where can you use Claude?
#Chat and apps
The web app, plus iOS, Android and desktop clients for macOS, Windows and Linux. All download routes are collected on the download page.
#In the terminal
Claude Code runs agentic coding sessions from your shell, with skills, subagents, hooks and plugins. Included on every paid plan. See Claude for developers.
#In your own product
The Claude API, plus AWS Bedrock, Google Cloud Vertex AI and Microsoft Foundry. Python and TypeScript SDKs, or plain HTTP.
Beyond chat, Anthropic now ships several task-specific surfaces: Cowork for agentic file and task work without a terminal, Claude in Chrome for browsing, Claude for Excel for spreadsheets, and the preview-stage Claude Design and Claude Science. Which of these your plan unlocks is laid out in the feature and plan matrix.
#Connecting Claude to your own tools
The mechanism is the Model Context Protocol — an open standard, originally from Anthropic, donated to the Linux Foundation's Agentic AI Foundation in December 2025 and now also adopted by OpenAI and Block. An MCP server exposes tools and data; Claude calls them. Anthropic documents connectors for Notion, Asana, Sentry, Stripe, HubSpot, GitHub, Airtable and Zoom, among others.
If you want Claude touching real systems, start with MCP and Claude agents, then automation patterns for the orchestration and cost-control side. If you would rather assemble workflows visually, Claude in n8n covers the node-based route.
#Is Claude output safe to use commercially?
Yes, on every plan including Free. Anthropic's Consumer Terms assign users “all of our right, title, and interest—if any—in Outputs”, and the Commercial Terms state that the customer owns its outputs. The single universal restriction is that you may not use outputs to train a competing model. Business customers get a stronger promise still: the Commercial Terms say Anthropic may not train on customer content.
Consumer plans are different — since the policy change of 28 August 2025, training on consumer chats is opt-in, with five-year retention if you allow it and 30 days if you decline. The detail, with the clause language, is on commercial use rights.
#Start here
#Frequently asked questions
What is the latest Claude model in August 2026?
Claude Opus 5, released 24 July 2026, is Anthropic's newest general flagship. Claude Fable 5, generally available since 9 June 2026, is positioned above it for long-running agent work and costs twice as much. Sonnet 5 arrived 30 June 2026 and remains the recommended default for most workloads.
Is Claude free to use?
Yes. The free plan includes web search, memory, file creation, code execution, one custom connector, extended thinking, Skills and up to five Projects. It does not include Claude Code, Cowork, Research, chat search, the Chrome extension, model selection or the 1M-token context window.
How many messages can I send on Claude Pro?
Anthropic does not publish an absolute number. Usage is limited by a rolling five-hour session window plus a weekly cap that resets on a fixed day assigned to your account, and Opus models have their own separate weekly counter. Published figures are relative multipliers only.
Does Claude have an app?
Yes — official apps exist for iOS, Android and ChromeOS, plus desktop apps for macOS 11 or later, Windows 10 or later and a Linux beta. There is also a Chrome extension and Microsoft 365 add-ins. All download routes are listed on our download page.
Can Claude generate images or video?
No. As of August 2026 Claude generates text and code only. It cannot produce images, video, audio or music, and it cannot accept video or audio as input. It does read images and documents. For generative media you need ChatGPT or Gemini.
Is claudiai.com the official Claude website?
No. This is an independent reference site with no affiliation to Anthropic. The official product lives at claude.ai, the company at anthropic.com, and developer documentation at platform.claude.com. We link to all three rather than reproducing account or payment functions.
Model specifications checked against platform.claude.com/docs and plan prices against claude.com/pricing on 21 August 2026. Anthropic changes models, prices and limits without notice; re-check both before making a purchasing or architectural decision.