Claude is the AI assistant built by Anthropic, and the strongest competitor to ChatGPT on frontier reasoning, coding, and long-form writing. Since its public debut in March 2023 alongside GPT-4, Claude has built a reputation for thoughtfulness, safety, and honesty — a product that often behaves like a careful colleague instead of an eager generalist, which makes it the default choice for a lot of professional work even when it's not the flashiest option on the market.
This page is the living reference to Claude — the model family, pricing, what's new, how it compares, and every story we've published. We update it whenever Anthropic ships a new model, changes pricing, or generates news that affects how Claude gets used.
As of April 2026, Claude Opus 4.7 is the current frontier model, with an 82% pass rate on SWE-bench Verified and the strongest agentic reliability in the industry. Sonnet 4.5, released in late 2025, remains the daily-driver default for most users — faster and cheaper, and on many benchmarks it beats the older Opus models it succeeded. This "Sonnet catches Opus" pattern has held across every model generation and is now the expected dynamic.
We rated Claude 4.8 / 5. The thoughtful person's AI chatbot. If you write, code, or do nuanced reasoning for a living, Claude is the model that respects your work — better at long-form output and instruction-following than anything else available. Pro at $20/month is a no-brainer.
Read the full Claude review →Latest news about Claude
Every story we've published tagged Claude. Refreshed automatically every time the newsroom publishes.
- Aug 25, 2026Anthropic merges Claude and Claude Cowork memory into one systemClaude will now carry context across chat and Cowork, and users can read, edit, or delete stored memories on any topic.
- Aug 24, 2026Stanford: AI is cutting entry-level jobs, with a 19% employment gap for young workersEmployment for 22-to-25-year-olds in AI-exposed jobs has fallen 11% since 2022, while less-exposed peers grew 10%.
- Aug 21, 2026OpenAI claws back ground on Anthropic among US businesses, Ramp data showsAnthropic still leads at nearly 44% share to OpenAI's 40%, but the ChatGPT maker is now growing faster in Q3 among Ramp's business customers.
- Aug 19, 2026Anthropic's Claude watermarks broken within four hours of launchDeveloper Guillaume Meyer's override went viral on GitHub with 20,000 X bookmarks and 100+ contributors, exposing the fragility of EU-mandated AI watermarking.
- Aug 18, 2026Anthropic tells investors revenue run rate hit $65B in JulyA sevenfold jump in a year puts Anthropic ahead of OpenAI's $40B run rate as it preps a confidential IPO at a $965B valuation.
- Aug 18, 2026Stanford's AI Observatory finds Anthropic filters out 48% of Claude conversationsA new independent dataset shows sensitive AI use — companionship, health, harassment — runs far higher than company reports admit.
Timeline of major releases
- Mar 14, 2023Claude launchesAnthropic's first public model drops alongside GPT-4. Positioned from day one on helpfulness, harmlessness, and honesty, trained with Constitutional AI.
- Jul 11, 2023Claude 2 and the 100K context windowFirst major model jump. The 100K token context window was unmatched at the time and made Claude the default for summarizing long documents.
- Mar 4, 2024Claude 3 family (Haiku, Sonnet, Opus)Three-tier model family mirroring OpenAI's. Opus briefly beats GPT-4 on several benchmarks and takes mindshare for coding and long-form writing.
- Jun 20, 2024Claude 3.5 Sonnet and ArtifactsA smaller, cheaper Sonnet beats the larger Opus on most tasks. Artifacts — a side pane where Claude renders live code, documents, and interactive components — becomes a hallmark feature.
- Oct 22, 2024Claude 3.5 Sonnet (new) + Computer UseFirst frontier model with native computer use (clicking, typing, scrolling on the user's behalf) via a developer API. The feature ships in public beta.
- May 22, 2025Claude 4 family (Sonnet + Opus)Full 4-series launch with longer context (up to 1M tokens in Sonnet 4 API), improved coding, and the "extended thinking" mode as a first-class feature.
- Sep 29, 2025Claude Sonnet 4.5Best coding model on public benchmarks at release. Anthropic's pattern of "Sonnet beats Opus" becomes the rule, not the exception.
- Mar 10, 2026Claude Opus 4.6 launchNew frontier Opus with sharper agentic reliability and longer autonomous task horizons.
- Apr 19, 2026Claude Opus 4.7 ships82% pass rate on SWE-bench Verified, a 3-point jump from 4.6. New "think" mode via budget_tokens parameter. Pricing unchanged at $15/$75 per million tokens.
How to access Claude
| Plan | Price | Includes |
|---|---|---|
| Free | $0 | Claude Sonnet with daily message caps. Artifacts and file uploads included. |
| Pro | $20/mo | 5x more usage, access to Opus and Sonnet, Projects, Claude Code, priority access to new features. |
| Max | $100–$200/mo | Highest usage caps, priority on new models, dedicated support. Two tiers (5x and 20x Pro limits). |
| Team | $30/user/mo | Pro features plus shared Projects, admin console, centralized billing, minimum 5 seats. |
| Enterprise | Contact sales | SSO/SCIM, audit logs, SOC 2 Type II, custom data residency, dedicated support. |
Claude vs. ChatGPT vs. Gemini
| Feature | Claude | ChatGPT | Gemini |
|---|---|---|---|
| Largest context window (paid) | 200K (1M in API) | 128K | 1M+ (2M in Advanced) |
| Native code tool | Claude Code + Artifacts | Sandbox + Codex CLI | Canvas |
| Real-time voice | No | Advanced Voice | Gemini Live |
| Image generation | No (text and code only) | Yes (DALL-E / GPT Image) | Yes (Imagen) |
| Computer use | Yes (beta) | Agent mode (GA) | Project Astra (limited) |
| Ad-free commitment | Yes (stated April 2026) | No public commitment | No (ads in some surfaces) |
| Training on user data | No by default | Yes unless opted out (Free/Plus) | Yes unless opted out |
The Claude family
Anthropic ships three model sizes per generation: Haiku (fastest, cheapest), Sonnet (default), and Opus (most capable). The tradeoff is roughly: Opus is ~5x the cost of Sonnet for incremental quality gains that matter mostly on hard reasoning and long-horizon agent tasks. For daily use — drafting, summarizing, coding at human-supervised pace — Sonnet is almost always the right default.
The family has been versioned in point releases rather than big-bang announcements: Claude 3, 3.5, 4, 4.5, 4.6, 4.7. Anthropic's cadence is quietly relentless — by our count, there has been a new frontier model roughly every four months since 2023, each one measurably better than the last on the benchmarks that matter most to power users (coding, long-context retrieval, agent task completion).
Artifacts and Claude Code
Artifacts is the feature that distinguishes Claude from most competitors' chat UIs. When Claude writes code, a document, or an interactive component, it appears in a side pane and runs live — click a button, edit a paragraph, iterate on the design without re-generating from scratch. It's a better fit for creative work than the linear chat transcript that still dominates ChatGPT and Gemini, and it's the reason a lot of writers and designers prefer Claude for anything longer than a paragraph.
Claude Code is a separate product, aimed at developers. It runs in the terminal, edits files directly in your codebase, runs tests, and commits to git. Unlike ChatGPT's web-based code interpreter, Claude Code lives on your machine and has the context of your real project. It has become Anthropic's flagship developer product in 2026 — used by solo indie developers and large engineering teams alike, and increasingly the reason teams pick Claude over ChatGPT for programming work. See our AI agents guide for how Claude Code fits into autonomous developer workflows.
Pricing and plans
Claude's consumer pricing mirrors ChatGPT's: free tier with caps, $20/month Pro, higher-tier Max at $100 or $200/month for heavy users, Team and Enterprise for organizations. The pricing has been stable since Pro launched in September 2023, following the same discipline we see at Anthropic across most decisions: raise capabilities, hold prices.
API pricing for Opus has held at $15 per million input tokens and $75 per million output tokens since Claude 3 Opus shipped in March 2024 — through three frontier model upgrades. This matters because the cost per useful unit of work has effectively dropped three times without a sticker-price change.
Privacy and training
Anthropic doesn't train on consumer conversations by default, which is a meaningful differentiator from OpenAI and Google. API traffic is also not used for training. Enterprise customers can negotiate additional data handling commitments, including zero-retention agreements where prompts and responses aren't stored beyond the session.
Claude is also the first major AI product to publicly commit to staying ad-free — a pledge made explicit in an Anthropic blog post in April 2026, and one that positions Claude against the competitive drift toward ad-supported free tiers. See our coverage of Claude's ad-free commitment for context.
Safety and alignment
Anthropic was founded with safety as a stated priority, and Claude's behavior reflects that — it's more cautious on persuasion, impersonation, dual-use content, and high-stakes advice than its competitors. The Constitutional AI training method, which has Claude critique and revise its own outputs against a written set of principles, is visible in practice: Claude will explain its reasoning, flag uncertainty, and decline more ambiguous requests than ChatGPT or Gemini.
The Responsible Scaling Policy is Anthropic's public commitment to specific safety evaluations at capability thresholds — AI Safety Level 2 (current), AI Safety Level 3 (coming), and higher levels that would trigger deployment restrictions. This is how Anthropic approaches the bet that models will keep getting dramatically more capable, and it's the most concrete version of a "stop here until we can prove this is safe" policy that any major AI lab has published. Our AI safety and alignment guide goes deeper.
Who should use Claude
Claude is the strongest default for professional writing, coding, and anything that benefits from careful nuance. If you're a developer, Claude Code is worth trying even if you're happy with GitHub Copilot — the context awareness and autonomy are different categories. If you're a writer, Artifacts changes the feel of working with an AI. If you're a business, Anthropic's API is the right choice when you need long context windows (up to 1M tokens in Sonnet 4) or when privacy is non-negotiable.
Claude is weaker than ChatGPT for real-time voice, native image generation, and anything involving casual consumer surfaces (DALL-E is integrated into ChatGPT in ways Claude doesn't match). Gemini remains ahead for long-video understanding and Google Workspace integration. See our full AI tools comparison for side-by-side task-specific guidance.
Our coverage
Every story we've tagged Claude appears above in Latest news. For the company, see our Anthropic hub. For head-to-head alternatives, see ChatGPT and Gemini. For the deeper strategy and safety questions, see our AI safety and alignment guide.
