Claude

Overview

Claude is Anthropic’s AI assistant for chat, coding, research, and agent-style work. You use it at claude.ai (web, desktop, iOS, Android), through product surfaces like Claude Code, and via the Claude API / Claude Platform for production apps.

Anthropic positions Claude as a “thinking partner” for long, hard tasks—writing, analysis, coding, and multi-step work with tools—rather than a pure chatbot. The consumer product and the API share the same model family (Haiku, Sonnet, Opus, Fable, and limited Mythos) but different billing: flat subscriptions for chat products versus pay-per-token for the API.

As of mid-2026, the main individual plans are Free, Pro ($20/mo or $17/mo annual), and Max (from $100/mo for 5× Pro usage, or $200 for 20×). Teams get Standard and Premium seats; Enterprise adds seat + usage-at-API-rates with admin/compliance controls. Claude Code, Cowork, Design, and Science are included on paid consumer and team plans (subject to shared usage pools).

Key features

• Chat everywhere — Web, desktop (macOS/Windows), iOS, and Android with the same account and shared usage pool.
• Model ladder — Haiku 4.5 (fast/cheap), Sonnet 5 (default workhorse; intro API pricing through Aug 31, 2026), Opus 4.8 (heavy reasoning/agents), Fable 5 (flagship long-running work; capacity-gated on subscriptions), Mythos 5 (limited availability).
• Extended thinking — Deeper internal reasoning for hard multi-step problems (consumes more of your usage allotment).
• Projects — Workspaces with persistent instructions and document knowledge; project content is cached so reuse burns less of your message budget. Paid plans get unlimited projects and RAG-style project knowledge modes.
• Artifacts — Side-panel creations (docs, tools, visualizations, interactive apps) you can iterate and share; persistent artifact storage on Pro+.
• Memory & chat search — Cross-chat memory and (on paid plans) search over past conversations so you do not re-paste the same context.
• Research — Multi-step research with web sources and citable outputs (paid plans).
• Web search, connectors & MCP — Live web search; connectors for Google Workspace, Slack, Microsoft 365, and remote MCP tools for custom context.
• Code execution & files — Create/edit files and run code in a sandbox for analysis and generation workflows.
• Claude Code — Terminal/agent coding included on paid plans; shares the same five-hour and weekly usage pool as chat.
• Claude Cowork, Design, Science — Task-delegation, design, and science-oriented product modes on Pro+ / Team (as listed on Anthropic’s pricing page).
• Claude API / Platform — Messages API, prompt caching, Batch API (50% off), computer use, Managed Agents ($0.08 per session-hour + tokens), web search ($10 / 1K searches), and cloud availability on AWS, Google Cloud, and Microsoft Foundry.

Pricing

Prices below are USD, exclude tax, and follow Anthropic’s published plans as of July 2026. Always re-check claude.com/pricing and the Platform docs before budgeting—promotions and capacity rules change.

Individual (claude.ai)

Plan	Price	What you get (summary)
Free	$0	Chat on web/mobile/desktop; web search; memory; files & code execution; connectors/MCP; extended thinking. Tight usage limits.
Pro	$20/mo, or $17/mo billed annually ($200 upfront)	At least ~5× Free usage per 5-hour session; Claude Code, Cowork, Design, Science; unlimited projects; Research; more models; Claude for Microsoft 365.
Max 5×	$100/mo	Everything in Pro + ~5× Pro usage per 5-hour session; higher output limits; priority access; early advanced features.
Max 20×	$200/mo	Same as Max 5× but ~20× Pro usage per session—aimed at all-day power users and heavy Claude Code work.

Team & Enterprise

• Team Standard — $25 per seat/month monthly, or $20/seat with annual billing. More usage than Pro; SSO, central billing, Claude Code/Cowork/Design/Science, enterprise search, no training on content by default.
• Team Premium — $125 per seat/month monthly, or $100/seat annual. About 5× Standard seat usage; mix Standard and Premium seats in one org.
• Enterprise — From ~$20/seat plus usage at API rates (self-serve and sales-assisted). SCIM, RBAC, audit logs, Compliance API, custom retention, IP allowlisting, HIPAA-ready option, spend controls. Annual contracts for sales-assisted paths.
• Education — University-wide plans with discounted student/faculty access (contact Anthropic).

API / Claude Platform (per million tokens)

Model	Input	Output	Notes
Haiku 4.5	$1	$5	High-volume, latency-sensitive
Sonnet 5	$2 intro / $3 standard	$10 intro / $15 standard	Intro through Aug 31, 2026; $3/$15 from Sep 1, 2026
Opus 4.8 (and 4.5–4.7)	$5	$25	Complex agents & coding; 1M context at standard rates
Fable 5 / Mythos 5	$10	$50	Top-tier long-running intelligence; Mythos limited

Cost cutters: prompt caching (cache reads ~0.1× input), Batch API (50% off input/output), and model routing (Haiku vs Sonnet vs Opus). Extras include web search at $10 per 1,000 searches, Managed Agents session runtime at $0.08/hour plus tokens, Opus fast mode at 2× standard (Opus 4.8: $10/$50), and US-only inference at 1.1× on supported models. Code execution includes a large free monthly container allotment then ~$0.05/hour.

Limits & gotchas

• No fixed message counts. Limits depend on model, message length, attachments, tools (Research, search), effort/thinking, and artifacts. Anthropic uses rolling 5-hour session windows plus weekly caps on paid plans.
• One shared pool. Web, desktop, mobile, and Claude Code all draw from the same plan usage. A long Code session can empty chat for the rest of the window.
• Pro runs out for power users. Community reports (HN/Reddit) consistently say $20 Pro is fine for light–moderate chat but burns quickly on Opus + large codebases; Max 5×/20× is the real daily-driver for heavy coding.
• Capacity-gated models. Fable 5 access has been staged with temporary promotions. As of mid-July 2026 announcements, Max and Team Premium keep included Fable at reduced (e.g. 50%) limits; Pro/Team Standard shift toward usage credits (with one-time credits in some rollouts). Treat top-model access as capacity-dependent, not permanent entitlements.
• Weekly Code promotions end. Temporary +50% Claude Code weekly limit boosts (e.g. May–July 2026) expire—plan capacity drops when promos end.
• API tokenizer note. Opus 4.7+, Fable 5, Mythos, and Sonnet 5 use a newer tokenizer that can produce ~30% more tokens for the same text—budget accordingly vs older Sonnet 4.6-class models.
• Programmatic policy. Anthropic has restricted using subscription/Claude Code harnesses to build competing CLIs or to route unlimited “subscription-priced” automation. Serious automation should use Console API keys and rate limits you control.
• Privacy modes differ. Consumer plans generally allow opt-out of training; Team/Enterprise default to no training on org content. API data handling follows commercial terms—review before regulated workloads.
• Refunds are limited outside statutory windows (e.g. EU 14-day); App Store/Play billing is handled by Apple/Google.

Community sentiment

On Reddit and Hacker News through 2026, Claude’s reputation is strong on writing quality, long-document reasoning, and coding agents, with recurring friction around usage limits and pricing clarity.

• Loved: Careful prose and instruction following; Projects + Artifacts for iterative work; Claude Code as a serious terminal agent; API caching/batch for production cost control; safety posture for enterprise evals.
• Complained about: Hitting 5-hour limits mid-flow (especially Max users running multi-agent Code); Pro feeling like a “gateway” to $100–$200 Max; Fable capacity churn (included → credits → partial re-inclusion); opaque “how many messages left” compared to fixed quotas; comparison shoppers who prefer ChatGPT Pro’s bundled image/Codex stack at the same $200 tier.
• API vs sub: Power users often note that raw API spend for the same Code volume can be many times a Max subscription—hence the loyalty to Max when interactive work fits the caps, and the migration to API when automation exceeds them.

“Pro is the $20 gateway; for real Claude Code days you plan around Max sessions or API credits—not infinite prompts.” — common HN/r/ClaudeAI theme, 2026

Who should use it

• Writers, analysts, PMs, and researchers who live in long documents and want Projects + Research + clean prose.
• Software engineers using Claude Code daily who can budget Max 5×/20× or pair Pro with careful session hygiene.
• Product teams standardizing on Team seats for SSO, shared connectors, and admin controls without full enterprise procurement.
• Builders shipping agents who need Messages API, tool use, prompt caching, Batch, or Managed Agents with predictable token economics.
• Regulated / enterprise buyers needing SCIM, audit logs, HIPAA-ready options, and no-training-by-default org policies.

Skip or dual-home if you primarily need native image generation, a broader free tier, or multi-model shopping in one UI—those workflows often sit better on ChatGPT or Gemini alongside Claude.

Alternatives

• ChatGPT — Strong all-rounder with image gen, GPTs, and OpenAI’s Pro-tier coding stack; better if you want one OpenAI bill for chat + media.
• Gemini — Google’s assistant with deep Workspace/Drive ties and competitive multimodal features.
• Perplexity — Search-first answers with citations when research UX is the main job.
• Claude Code — Anthropic’s coding agent surface (also covered under paid Claude plans); compare vs IDE agents if coding is 100% of the use case.
• Cursor — AI-native IDE that can route multiple models; choose when the editor, not the chat site, is home base.
• DeepSeek / Mistral — Cost-sensitive API alternatives for bulk inference where Anthropic quality premium is optional.
• GitHub Copilot — IDE-centric completion and chat deeply integrated with GitHub repos and PRs.

Verdict

Claude in 2026 is a mature dual product: a polished subscription assistant (Free → Pro → Max → Team/Enterprise) and a full developer platform (Haiku/Sonnet/Opus/Fable with caching, batch, and agents). It earns its reputation on hard reasoning, writing, and coding agents—but you must design around shared 5-hour and weekly limits, capacity-gated flagship models, and the split between subsidized interactive use and true API scale.

Bottom line: Start Free or Pro to learn the workflow; move to Max if Claude Code and long sessions are daily; use Team for org admin; put production automation on the API with caching and the right model tier. Re-verify Fable/Opus entitlements and Sonnet 5 intro pricing dates on Anthropic’s site—those numbers move with capacity and promotions.