Claude Opus 4.8 2026: $380B Valuation, 69.2% SWE-Bench Pro
Last updated: 2026-06-14
Claude by Anthropic now runs on Opus 4.8 (69.2% SWE-Bench Pro, 83.4% OSWorld). Anthropic is valued at $380B. Plans range free to $200/mo Max with 1M context.
Claude, built by Anthropic (founded 2021, San Francisco), is a frontier AI model family whose latest flagship, Opus 4.8, was released May 28, 2026 and scores 69.2% on SWE-Bench Pro and 83.4% on OSWorld-Verified. Anthropic's $30 billion Series G in February 2026 valued the company at $380 billion, and Claude Code alone reached $2.5 billion in annualized revenue that month. Plans start free, with Pro at $20/month, Team at $20-25/seat/month, and Max at $100-200/month for a 1 million token context window.
About Claude
Claude is an AI assistant built by Anthropic, a public benefit corporation founded in 2021 by Dario Amodei (CEO) and Daniela Amodei (President), both former senior leaders at OpenAI. As of early 2026, claude.ai serves 19 million monthly active web users, receives 287 million site visits per month, and supports over 300,000 business customers. Anthropic raised $30 billion in its Series G in February 2026 at a $380 billion valuation, and reports place annualized revenue at $30 billion as of April 2026. The current flagship model is Claude Opus 4.8, released May 28, 2026, scoring 88.6% on SWE-bench Verified and 93.6% on GPQA Diamond. Claude's most defensible technical advantage is context capacity. The Opus family supports 1 million input tokens (approximately 750,000 words or a 3,000-page document), the largest context window available commercially in 2026. This makes Claude the preferred model for legal document review, long-form analysis, large codebase understanding, and research synthesis tasks where competitors must chunk inputs. Claude was trained with Constitutional AI, a Anthropic-developed technique that has the model evaluate and improve its own outputs against a set of principles, making it notably less prone to harmful outputs and more consistent at nuanced instruction-following than other frontier models. The 2026 feature set spans three areas. In personal productivity: Projects (persistent workspaces with shared memory), Artifacts (generated code, documents, and diagrams saved as reusable outputs), and Cowork (real-time collaborative sessions). In agentic capabilities: Claude Code (a terminal-native coding agent that reads entire codebases, edits files, runs tests, and commits to git), Computer Use (the AI operates desktop applications via screenshot-action cycles), and Skills (pre-built agent workflows). In platform integrations: native MCP (Model Context Protocol) support, Microsoft 365 Copilot integration, Amazon Bedrock, Google Vertex AI, GitHub Copilot (as underlying model), and Cursor IDE. Pricing runs from free (base models) to Pro at $20/month, Max at $100/month (5x capacity) or $200/month (20x capacity), Team at $25/seat/month ($20 annual), and Enterprise at custom pricing with BAA for HIPAA-eligible use cases. The API charges per token: Haiku 4.5 at $1/$5, Sonnet 4.6 at $3/$15, and Opus 4.7 at $5/$25 per million input/output tokens. Batch processing cuts all rates by 50%. Claude's primary weakness is the absence of native image, audio, or video generation. It accepts image and document inputs but outputs only text and code. For generative media tasks, users must switch to DALL-E, Midjourney, or Runway. The Trustpilot rating reflects a wave of April 2026 billing complaints tied to unauthorized gift-subscription charges, not the model quality, which consistently earns top marks from professional users on G2 and Capterra.
Screenshots
Pricing
Free tier with base model access. Pro at $20/month. Max at $100/month (5x capacity) or $200/month (20x capacity). Team at $25/seat/month ($20/seat annual), 5-seat minimum. Enterprise custom-priced. API: Haiku 4.5 at $1/$5 per million tokens, Sonnet 4.6 at $3/$15, Opus 4.7 at $5/$25. Batch processing saves 50% on all API rates. Cache reads at ~10% of input rate.
Feature Comparison by Tier
| Feature | Pro | Free | Max 5x | Max 20x |
|---|---|---|---|---|
| Usage vs Free | 5x | 1x | 25x | 100x |
| Default model | Sonnet 4.6 | Sonnet 4.6 | Opus 4.6 | Opus 4.6 |
| Context window | 200K | 200K | 1M | 1M |
| Extended thinking | yes | - | yes | yes |
| Claude Code | yes | - | yes | yes |
| Claude Design | yes | - | yes | yes |
| App connectors | 20 apps | 20 apps | 200+ apps | 200+ apps |
| Priority queue | - | - | - | yes |
Key Features
- 1 Million Token Context Window: Claude Opus supports 1 million input tokens (approximately 750,000 words), enabling full codebase analysis, 3,000-page document review, and multi-source research synthesis in a single session.
- Claude Code (Agentic Coding): Terminal-native coding agent that reads entire codebases, writes and edits files across multiple directories, runs tests, and commits to git, reaching $2.5 billion in annualized revenue by February 2026.
- Extended Thinking and Constitutional AI: Claude's Constitutional AI training gives it structured self-evaluation on outputs, making it the strongest frontier model for nuanced instruction-following, safety-critical tasks, and long-form reasoning in 2026.
- Projects and Artifacts: Persistent workspaces (Projects) retain memory and context across sessions; Artifacts save generated code, documents, and diagrams as reusable outputs that teammates can reference and build on. Claude Code now supports shareable live Artifacts with automatic updates and version history.
- Voice Mode and Mobile Integration: Voice conversations are free on iOS and Android with five voice options; iOS supports camera input, Siri Shortcuts, and file sharing from any app via the share sheet.
- MCP and Platform Integrations: Native MCP (Model Context Protocol) support connects Claude to thousands of external tools. Admins can provision enterprise MCP connectors (including Okta) once, and users get zero-touch access across Claude Chat, Code, and Cowork. Claude is also the underlying model for GitHub Copilot, Cursor IDE, and Microsoft 365 Copilot on enterprise plans.
- Computer Use: Claude can operate desktop applications autonomously by interpreting screenshots and executing mouse clicks and keyboard actions, enabling end-to-end task automation without custom integrations.
- Frontier Benchmark Performance: Claude Opus 4.8 scores 88.6% on SWE-bench Verified and 93.6% on GPQA Diamond, leading public alternatives on both software engineering and graduate-level science. Fable 5 (June 2026), Anthropic's new Mythos-class model, expands the lineup with a focus on safety and broad general use.
Pros
- 1 million token context window on Opus is 8x larger than ChatGPT-4o's 128K window, making Claude the only frontier model capable of processing an entire large codebase or legal document in one session.
- Claude Opus 4.8 leads SWE-bench Verified at 88.6% and GPQA Diamond at 93.6%, ahead of all publicly benchmarked alternatives as of May 2026.
- SOC 2 Type II, ISO 27001:2022, HIPAA (with BAA), and Zero Data Retention make Claude the compliance-ready choice for regulated industries including healthcare, legal, and financial services.
Cons
- No native image, audio, or video generation: Claude accepts image inputs but outputs only text and code, requiring a separate tool for any generative media task.
- Trustpilot shows a wave of April 2026 billing complaints over unauthorized gift-subscription charges and double-billing, indicating Anthropic's billing and support operations have not kept pace with growth.
- Free tier is limited to base models (Haiku); Pro at $20/month is needed to access Sonnet 4.6 and occasional Opus usage, making the true cost of the full capability higher than the headline free tier implies.
Product Information
- Cloud
- Yes
- Self-Hosted
- No
- On-Premise
- No
- Languages
- English, French, German, Spanish, Portuguese, Japanese, Chinese (Simplified), Chinese (Traditional), Korean, Italian
- Training
- Documentation, Help Center, API Reference, Video tutorials, Community Forum
Frequently Asked Questions
What is Claude and who built it?
Claude is a family of AI models built by Anthropic, founded in 2021 by Dario and Daniela Amodei, former OpenAI executives, and based in San Francisco. The current flagship, Claude Opus 4.8, was released on May 28, 2026, and leads on benchmarks including 69.2% on SWE-Bench Pro, 83.4% on OSWorld-Verified, and 57.9% on Humanity's Last Exam. Anthropic raised a $30 billion Series G round in February 2026 at a $380 billion valuation, with investors including Google, Amazon, Spark Capital, Salesforce Ventures, and Menlo Ventures. Anthropic's annualized revenue reached $14 billion in February 2026, up 55.56% year over year, and Claude Code alone reached $2.5 billion in run-rate revenue by the same month. Claude is available via claude.ai, mobile apps, the API, and as the underlying model for GitHub Copilot, Cursor, and Microsoft 365 Copilot on enterprise plans.
How much does Claude cost?
Claude has six pricing tiers. The Free tier is $0/month with limited access to Claude Haiku 4.5 and Sonnet 4.6. Pro costs $20/month with a 200K token context window and primary access to Sonnet 4.6 plus limited Opus 4.7. Max comes in two levels: $100/month for 5x the Pro capacity (with access to Opus 4.7/4.8 and a 1 million token context window) or $200/month for 20x capacity with priority access to Opus 4.8. Team costs $25/seat/month billed monthly or $20/seat/month billed annually, with a 5-seat minimum. Enterprise is custom-priced. On the API, Claude Haiku 4.5 costs $1/$5 per million input/output tokens, Sonnet 4.6 costs $3/$15, and Opus 4.8 starts at $5/$25 (with a $10/$50 fast mode running at 2.5x speed). Batch processing cuts API costs by 50%, and cached reads cost roughly 10% of the standard input rate.
What does Claude do that competitors don't?
Claude Opus 4.8, released May 28, 2026, leads rival models on several benchmarks: 69.2% on SWE-Bench Pro, 83.4% on OSWorld-Verified, 93.6% on GPQA Diamond, and 57.9% on Humanity's Last Exam, and is roughly 4x less likely to let its own code flaws pass unflagged than prior versions. Claude Max plans offer a 1 million token context window, enabling full-codebase analysis and review of documents up to roughly 3,000 pages in a single session. Claude Code, its terminal-native coding agent, reached $2.5 billion in annualized revenue by February 2026, reflecting heavy adoption for autonomous, multi-file coding tasks that include running tests and committing to git. Claude is also the model underlying GitHub Copilot, Cursor IDE, and Microsoft 365 Copilot on enterprise plans, giving it reach beyond its own first-party apps. New Opus 4.8 features include Dynamic Workflows for parallel subagents and an effort-control slider for balancing speed against depth.
How does Claude compare to ChatGPT?
Claude and ChatGPT (OpenAI) lead in different areas. Claude is generally regarded as the stronger choice for complex coding, debugging, and large-codebase work, with Claude Opus 4.8 scoring 69.2% on SWE-Bench Pro and 83.4% on OSWorld-Verified, and Claude Code generating $2.5 billion in annualized revenue by February 2026. ChatGPT leads on raw scale with 900 million weekly active users versus Claude's roughly 18.9-30 million monthly consumer users, and ChatGPT still leads on Terminal-Bench 2.1 (78.2% vs Opus 4.8's 74.6%). ChatGPT also offers native image and video generation (DALL-E, Sora) that Claude does not, while Claude's 1 million token context window on Max plans and Constitutional AI training give it an edge for long-document analysis and safety-critical tasks. Both offer a free tier and a $20/month entry-level paid plan, so the choice often comes down to coding depth (Claude) versus ecosystem breadth (ChatGPT).
Is Claude free / does it have a free tier?
Yes, Claude has a free tier at $0/month that includes limited access to Claude Haiku 4.5 and Sonnet 4.6, with voice mode free on iOS and Android. Free-tier usage caps reset daily and can fall back to a lighter model during peak hours. The free tier does not include Claude Opus 4.7/4.8, the 1 million token context window, or the higher usage limits of Max plans. Free consumer conversations on claude.ai are, by default, eligible for use in model improvement, with an opt-out available in settings; API and Enterprise customers have Zero Data Retention by default. For consistent access to frontier Opus models and significantly higher daily limits, users need Pro at $20/month or Max at $100-200/month.
Who is Claude best for / who should avoid it?
Claude is best for developers, researchers, and teams working with large codebases or long documents, thanks to its 1 million token context window on Max plans, Claude Code's terminal-native agentic coding, and frontier scores on SWE-Bench Pro (69.2%) and OSWorld-Verified (83.4%). Anthropic's Constitutional AI training and SOC 2 Type II, ISO 27001, ISO/IEC 42001, HIPAA, GDPR, and CCPA compliance make it a strong fit for regulated industries needing Zero Data Retention. Users who primarily want built-in image or video generation, the broadest third-party app ecosystem, or the largest possible user community may prefer ChatGPT, which has roughly 900 million weekly active users versus Claude's tens of millions. Heavy Claude Code users on Max plans should also be aware that Opus consumes about 3x the tokens of Sonnet per request, so budget-conscious users should default to Sonnet 4.6 for routine work.
Does Claude support tool use, MCP, and computer use?
Yes. Claude has native support for the Model Context Protocol (MCP), an open standard that connects Claude to thousands of external tools, databases, and services across desktop, web, and API integrations. Claude can also use Computer Use to operate desktop applications autonomously by interpreting screenshots and executing mouse clicks and keyboard actions, enabling end-to-end automation without custom integrations for each app. Claude Code, the terminal-native coding agent, can read entire codebases, edit files across multiple directories, run tests, and commit to git autonomously. Opus 4.8 added Dynamic Workflows, which run parallel subagents on different parts of a task, and an effort-control slider to trade off speed against depth of reasoning. If you run many MCP connectors simultaneously, disconnect the ones you're not using for a given session, since each adds schema overhead that can accelerate rate-limit consumption.
Does Claude train on user data and what compliance certifications does it have?
Anthropic does not train on API or Enterprise customer data, and both get Zero Data Retention by default. Free and Pro consumer users on claude.ai are opted into model improvement by default, with an opt-out available in account settings. Anthropic holds SOC 2 Type II, ISO 27001:2022, and ISO/IEC 42001:2023 certifications, the latter being a dedicated AI management system standard. Claude is HIPAA-compliant with a Business Associate Agreement (BAA) available on qualifying plans, and Anthropic complies with GDPR and CCPA. Since the EU AI Act's general-purpose AI (GPAI) obligations took effect in August 2025, Anthropic has been subject to those requirements, and its ISO 42001 certification provides a documented AI governance framework supporting that compliance. Organizations handling regulated data should use Enterprise or API access rather than the free consumer tier, where default training-data usage applies unless explicitly disabled.
Top Alternatives
- ChatGPT: Pick ChatGPT if you need built-in image generation, real-time web search, and OpenAI multimodal ecosystem.
- Gemini: Choose Gemini when your work lives inside Google Workspace and you need the tightest native Google integration.
- Grok: Pick Grok if you need live X (Twitter) data and real-time social context that Claude does not surface natively.