# Claude Claude is a [[Large Language Models (LLMs)]] created by [[Anthropic]]. There are multiple variants of Claude: - Haiku: small, fast and cheap ([[Claude Haiku 5.5]]) - Sonnet: the mid-tier workhorse ([[Claude Sonnet 5.5]]) - Opus: for complex work ([[Claude Opus 5.5]]) - Mythos class (Fable / Mythos): the top tier, above Opus ([[Claude Fable 5.1]]) Those evolve over time. As of mid-October 2026, the current lineup is the [[Claude 5|Claude 5.5 family]] plus Fable 5.1: - [[Claude Fable 5.1]] ($10 / $50 per million tokens) is the most powerful publicly available model, a point release of [[Claude Fable 5]]. Claude Mythos 5.1 shares its weights with lighter safeguards and is restricted to verified cybersecurity and life sciences organizations - [[Claude Opus 5.5]] ($4 / $20, released 2026-09-22) gets Fable 5.1-level results on most work. Anthropic recommends it for complex work that needs careful judgment. It's the default model in [[Claude Code]] - [[Claude Sonnet 5.5]] ($2 / $10, released 2026-09-28) is close to Opus 5.5 on many benchmarks, for well-scoped everyday tasks. It also powers the free tier on claude.ai - [[Claude Haiku 5.5]] (from $0.10 / $0.50, released 2026-10-07) is the fastest and cheapest, built for classification, routing, extraction and subagent work. Its price goes up five times for prompts over 100K tokens All four have a 1M-token context window, 128K max output and a June 2026 knowledge cutoff. Cache reads cost 10% of the input price on Haiku 5.5, 5% on Opus 5.5 and Sonnet 5.5 (Sonnet's were halved on 2026-10-07), and 2.5% on Fable 5.1. See [[Claude 5]] for how the generation fits together (shared weights, safeguards routing, the export-control interruption), and [[AI Model Selection]] for how to pick a model. The direct competitors at the same price points are [[GPT-6 Sol]] for Sonnet 5.5, [[GPT-6 Astra]] for Opus 5.5, and [[GPT-6 Luna]] for Haiku 5.5. Note that Fable 5 and Mythos 5 were pulled worldwide on 2026-06-12 by a US export-control directive, and access was restored on 2026-07-01. See [[2026-06-17 The US Government Banned Claude Fable 5 and Mythos 5|Fable 5 ban]]. I personally consider Claude as superior to [[GPT4]] (all variants) and [[Gemini]]. There's a free and a different paid tiers. Since October 2026, Max and Team subscribers also get monthly API credits for the [[Claude Platform]]: $100 on Max 5x, $200 on Max 20x, and up to $500 pooled across users on Team. You claim them by picking an API organization under Settings → Billing. They can be used on any model, they don't roll over, and you can turn off auto-reload so that API requests stop when the balance runs out. [[Simon Willison]] calls them "really generous", and notes that [[OpenAI Codex]]'s subscription still works out better for heavy API users. ## One Claude: chat and Cowork merged (September 2026) On 2026-09-16, Anthropic announced that [[Claude Cowork]] and chat are merging into a single Claude, rolling out to Pro and Max over a few weeks. [[Claude Design]] is folded in too, so you can ask for a slide deck, a design or a doc without switching tabs or apps. The idea, according to [[Cat Wu]]: people didn't want to choose which Claude product to use for a task, and the models are now good enough to route for you. Based on the prompt, Claude decides whether to answer quickly or do deeper agentic work, and which kind of output fits. You can still stop it, redirect it, or control effort and approach. Claude keeps working on a task after you close the laptop, and asks when something is unclear. I like this. I'd rather let the model route a task than pick between the Chat, Cowork and Code tabs myself. ## Duet (October 2026) On 2026-10-21, Anthropic announced [[Claude Duet]], a proactive agent in research preview. It runs on [[Claude Opus 5.5]] in its own cloud computer, checks your connected apps every few hours, sends you a morning brief, and asks before it sends messages, pays, signs or deletes anything. It's a US-only waitlist for Pro and Max subscribers, with no extra price, but it shares your plan's usage limits with the rest of Claude ([[Claude Code]] included). Duet lives in its own app (iOS, Android and claude.ai on the web) during the preview, which sits oddly next to the September merge of chat and Cowork into one Claude. ## Projects Claude Projects enable creating a stable "context" for conversations with Claude. Within the context of a project, Claude will already know what you have discussed before. It's useful when working on a project, to avoid having to repeat the same information over and over (e.g., project goals, scope, current status, previously discussed ideas, etc). On a Claude Project page, it's possible to configure the "Project" knowledge, that is, the starting context for all discussions. The Project knowledge can be defined using: - Custom instructions (i.e., a prompt) - Content (e.g., documents, text, images, etc) Not to be confused with [[Claude Code Projects]] (September 2026), where one conversation orchestrates parallel cloud threads. ## Styles Claude supports selecting different output styles. Some are built in, but it's also possible to create custom styles. Here's an introduction to this feature:  ## Computer Use See [[Claude Computer use]]. ## Desktop ### Introduction Anthropic provides an official desktop client for Claude. Since November 2024, Claude Desktop has become MUCH more powerful, because it can be integrated with arbitrary data sources through Anthropic's new open standard called [[Model Context Protocol (MCP)]]. ### Feature previews Click on your name in the bottom left corner, then select "Feature Preview" to view and enable feature previews. ### Integrate with Google Drive Claude can access data stored in Google Drive. To connect your account, click on your name in the bottom left corner, then choose "Settings", then click on "Connect" in the "Connected accounts" section. After connecting your account, a Google Drive icon will appear below the chat box, allowing you to load recent Google Drive files, or files by URL. ### Configuration - On Windows, the configuration is located in `%appdata%/Claude` - On MacOS, it is under `~/Library/Application Support/Claude` For configuring MCP, please look at [[Model Context Protocol (MCP)]]. ## Code Claude also has a CLI called [[Claude Code]], available in the Claude desktop app too (Code tab). You extend it with [[Claude Code Plugins|plugins]], and since October 2026 with [[Claude Code Mods|mods]]: small JS/TS files that run inside Claude Code and can change how it behaves and looks. You can ask Claude to write one for you. ## Marketplace The [[Claude Marketplace]] (claude.com/marketplace) is Anthropic's storefront around Claude: more than 2,000 connectors and plugins to add, agents and products from partners to buy (Factory, Cursor, Vercel, CrowdStrike and others), and service partners from the Claude Partner Network. Purchases count toward your existing Anthropic commitment. Publishers submit plugins and mods through the Claude directory (claude.ai/directory/manage). ## Design [[Claude Design]] is Anthropic's Labs product for generating designs, prototypes, slides, and one-pagers through conversation, with handoff to [[Claude Code]] for implementation. Since 2026-10-08, Design, Docs and Slides are out of beta and available on every plan, Free included. The standalone site (claude.ai/design) closes on 2026-12-14. ## Dashboards and Motion (October 2026) On 2026-10-08, Anthropic added two artifact templates in beta: Claude Dashboards and Claude Motion. In Claude, an artifact is anything it makes for you that you'd show to someone else: a doc, a deck, a design, a dashboard, an animation. It opens next to the conversation, gets saved in the Artifacts tab, and you edit it by asking Claude or by hand. Claude Dashboards (Pro, Max, Team and Enterprise) connects to a data warehouse (Amazon Redshift, BigQuery, ClickHouse, Databricks, Snowflake) or to another connector such as Salesforce. You ask a question in plain language, Claude writes one SQL query per chart, and the dashboard stays current as the data changes. Click a number and you see the query behind it, plus when that chart was last refreshed. Anthropic pitches it for quick, exploratory questions ("how do this week's signups compare with last month's?"). For deeper analysis, you send the dashboard to Amplitude, Grafana, Hex, Mixpanel, Omni, Perplexity, PostHog or Sigma; Looker, Tableau, Databricks Genie and monday.com are announced as "coming soon". Claude Motion (Team and Enterprise only) turns a report, a chart or a set of notes into a short animated explainer for an all-hands, a board deck or customer onboarding. Claude writes the animation as code that moves your text, charts, shapes and images. No video generation model is involved, so there's no generated footage and no AI-generated people. Since it's code, you can change one word, one number or the timing and the rest stays as it was. You can then export an MP4 or open it in Adobe, Descript, HeyGen, Runway and a few other video tools. The launch video (narrated by Nate, from Anthropic's design team) uses a bike-share example. The operations team notices that the docks on the hill are empty by 8:15 every weekday, has Claude pull a year of ride data from the warehouse into a dashboard, then asks Claude to turn that dashboard into an animation for the next all-hands. The tweaks happen through comments left right on the animation ("hold on the empty dock a beat longer"), and colleagues can comment on it too. The query behind each chart matters most to me. A dashboard built by a model is only worth something if you can check what each number actually counts. Same logic for Motion: building animations as code means the numbers on screen are the numbers you gave it. On Enterprise plans, both are off by default (Organization settings > Artifacts); Motion is on by default on Team plans. Both count toward your plan's usage limits, and longer animations eat more of it. Not to be confused with [[Claude Code Artifacts]], the live pages that [[Claude Code]] publishes from a coding session. ## Science [[Claude Science]] is Anthropic's AI workbench for scientists (public beta since June 2026): a separate app that runs analyses on your own machines and clusters, with full provenance on every result. ## Managed Agents [[Claude Managed Agents]] is Anthropic's cloud-hosted agent harness — run Claude as an autonomous agent in a secure container without building the loop yourself. See also [[Anthropic CLI (ant)]]. ## Developer Platform Building software on top of Claude happens through the [[Claude Platform]]: the [[Claude API]], the [[Claude Console]] web dashboard, docs, a cookbook, and official SDKs. The cookbook (https://platform.claude.com/cookbook/) has working recipes for tool use, agent patterns, RAG, prompt caching, batch processing, extended thinking, and more. See [[Claude Platform]] for the highlights. For practical know-how, Anthropic runs claude.dev, a developer hub with "tips, tricks, and POVs from Anthropic's developers": articles on agents, engineering, playbooks, skills and tutorials (getting the most out of Opus 5.5, [[Claude Code effort levels|spending your effort in Claude Code]], Claude Code mods), plus videos with founders and engineers. ## Model Context Protocol (MCP) Support Claude Desktop has deep support for [[Model Context Protocol (MCP)]] servers. It acts as an MCP client, but also supports [[Desktop Extensions for MCP]], enabling easy installation of local MCP servers: ![[Claude - DesktopExts Support.png]] ## References - Official Website: https://claude.ai/ - Claude Downloads: https://claude.ai/download - Anthropic's Website: https://www.anthropic.com/ - Claude Projects: https://www.anthropic.com/news/projects - Claude Styles: https://www.anthropic.com/news/styles - Cowork and chat merge (Cat Wu, 2026-09-16): https://x.com/_catwu/status/2100260655312089562 - Claude Marketplace: https://claude.com/marketplace - Claude Duet product page: https://claude.com/product/claude-duet - Introducing Claude Duet (2026-10-21): https://claude.com/resources/articles/claude-duet - claude.dev (Anthropic developer hub): https://claude.dev - Getting started with Claude Code mods: https://claude.dev/blog/getting-started-with-claude-code-mods - Current lineup comparison (Claude Haiku 5.5 overview, platform docs): https://platform.claude.com/docs/en/models/haiku-5-5/overview - Introducing Claude Haiku 5.5 (Sonnet 5.5 cache reads, subscriber API credits): https://www.anthropic.com/claude-haiku-5-5 - Monthly API credits for Max and Team plans: https://support.claude.com/en/articles/17154008-monthly-api-credits-for-max-and-team-plans - [[Simon Willison]], "Claude Haiku 5.5" (2026-10-07): https://simonwillison.net/2026/Oct/7/claude-haiku-5-5/ - Introducing Claude Dashboards and Claude Motion (Claude YouTube channel, 2026-10-08): https://www.youtube.com/watch?v=en0GuyhieQk - Build live dashboards and animate explainers with Claude (Claude blog, 2026-10-08): https://claude.com/blog/dashboards-and-motion - Get started with Claude Dashboards (Claude Help Center): https://support.claude.com/en/articles/17454700-get-started-with-claude-dashboards - Get started with Claude Motion (Claude Help Center): https://support.claude.com/en/articles/17454997-get-started-with-claude-motion - What are artifacts and how do I use them? (Claude Help Center, plan availability table): https://support.claude.com/en/articles/17153992-what-are-artifacts-and-how-do-i-use-them A comparison of Custom [[ChatGPT]] GPTs and Claude Projects:  ## Related - [[Claude 5]] - [[Claude Fable 5.1]] - [[Claude Opus 5.5]] - [[Claude Sonnet 5.5]] - [[Claude Haiku 5.5]] - [[Anthropic]] - [[Claude Code]] - [[Claude Duet]] - [[Claude Platform]] - [[Claude API]] - [[AI Model Selection]] - [[Large Language Models (LLMs)]]