Anthropic Claude Release Timeline
Every model release, feature launch, research paper, and milestone in Anthropic's history. 113 events from 2021 to today.
Claude Opus 5 Released
Claude Opus 5 — a step-change for the Opus tier, powering long-running agents with major improvements in coding and professional work. 1M context, up to 128K output, adaptive thinking on by default. API pricing $5/$25 per MTok.
Claude Fable 5 Redeployed
Fable 5 returns globally (July 1), Anthropic's most capable model for the hardest reasoning and long-horizon agentic work, alongside a proposed industry-wide jailbreak severity scoring framework. 1M context, always-on adaptive thinking. API pricing $10/$50 per MTok.
Claude Sonnet 5 Released
Anthropic releases Claude Sonnet 5, delivering frontier performance across coding, agents, and professional work at scale. 1M-token context window and adaptive thinking. API pricing $3/$15 per MTok (intro $2/$10 through Aug 31, 2026).
Claude Sonnet 4.6 Released
Anthropic releases Claude Sonnet 4.6, priced the same as Claude Sonnet 4.5 ($3/$15 per million tokens), as of February 17, 2026. This continues Anthropic's strategy of iterative improvements without price increases at the mid-tier, maintaining competitive positioning in the developer market.
Prompt Caching Workspace-Level Isolation
Starting February 5, 2026, prompt caching on the Anthropic API uses workspace-level isolation instead of organization-level isolation. Caches are now isolated per workspace, ensuring data separation between workspaces within the same organization for improved security and compliance.
Claude in PowerPoint Launched
Anthropic launches Claude as an add-in for Microsoft PowerPoint, allowing users to create, edit, and enhance presentations using natural language directly within PowerPoint. This extends Anthropic's suite of workplace productivity integrations beyond Excel.
Claude Opus 4.6 Released
Anthropic releases Claude Opus 4.6 with improved coding skills and an agent team feature. As of February 20, 2026, Opus 4.6 holds the longest task-completion time horizon estimated by METR at 14.5 hours (50th percentile). Claude in PowerPoint launches alongside this model, allowing Claude to function as an add-in within PowerPoint presentations.
Anthropic Valuation Reaches ~$380B
As of February 2026, Anthropic's estimated valuation reaches approximately $380 billion, reflecting continued investor interest in the company's enterprise AI leadership. This is nearly double the $183 billion valuation from the September 2025 Series F, driven by rapid revenue growth and Claude Code's commercial success.
Enterprise Plan Available Self-Serve (No Sales Required)
Anthropic makes Enterprise plans purchasable directly on the website without requiring a Sales conversation. Previously only available to customers working through the Sales team. Self-serve Enterprise includes access to Claude, Claude Code, and Cowork in a single seat type.
GOV.UK AI Assistant Partnership with UK Government
Anthropic announces selection to develop and pilot a Claude-powered AI assistant for GOV.UK, in partnership with the UK Department for Science, Innovation and Technology (DSIT) and the Government Digital Service. Initial focus on employment support, helping job seekers with personalized career advice and access to training programs.
Teach For All AI Literacy Partnership
Anthropic partners with Teach For All to launch the AI Literacy & Creator Collective (LCC), a global training initiative bringing AI tools and education to educators across 63 countries. Over 100,000 teachers and alumni participate across Teach For All's network serving 1.5 million students in under-resourced schools.
Anthropic Economic Index Full Report Published
Anthropic releases the full Economic Index Report analyzing Claude usage data from November 2025. The report covers geographic patterns, labor market impact, job exposure rates, and AI's effect on different professional categories. Found 49% of jobs now use Claude for tasks, with data entry and radiologists showing highest exposure.
Accenture Strategic Partnership Announced
Anthropic and Accenture announce a multi-year strategic partnership to help enterprises deploy Claude at scale. The Accenture Anthropic Business Group is established, with approximately 30,000 Accenture professionals trained on Claude. The partnership targets CIOs for AI deployment across engineering organizations.
MCP Donated to Linux Foundation's Agentic AI Foundation
Anthropic donates the Model Context Protocol to the newly established Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation. The foundation is co-founded by Anthropic, Block, and OpenAI with support from Google, Microsoft, AWS, Cloudflare, and Bloomberg. By its first anniversary, MCP has 10,000+ active public servers and 97 million monthly SDK downloads.
Cowork Desktop App Launched
Anthropic launches Cowork, a desktop application for non-developers to automate file and task management using Claude. Cowork is included in the Enterprise and self-serve Enterprise plans, providing AI-powered desktop productivity capabilities without requiring technical expertise.
Claude Code Security Launched (Research Preview)
Anthropic launches Claude Code Security in a limited research preview for Enterprise and Team plan customers. The tool provides AI-driven code analysis that reasons like a security defender, surfaces validated findings with suggested patches, and requires human approval before applying fixes.
Claude Opus 4.5 Released
Anthropic releases Claude Opus 4.5, with main improvements in coding and workplace tasks such as producing spreadsheets. The release also introduces 'Infinite Chats,' a feature that eliminates context window limit errors by managing conversation history automatically.
Anthropic Opens First Asia-Pacific Office in Tokyo
Anthropic officially opens its first Asia-Pacific regional office in Tokyo, Japan, marking a significant international expansion milestone. This is part of Anthropic's strategy to grow its global footprint and international workforce, with plans to also recruit across India, Japan, Korea, and Australia.
Agent Skills API (Beta) Launched
Anthropic launches Agent Skills in beta, a new way to extend Claude's capabilities through organized folders of instructions, scripts, and resources loaded dynamically. Skills include pre-built capabilities for PowerPoint, Excel, Word, and PDF file handling, plus a Custom Skills API for uploading domain-specific expertise.
Claude in Excel Launched
Anthropic launches Claude as an add-in for Microsoft Excel, enabling AI-powered spreadsheet assistance. Claude in Excel supports native Excel operations including pivot table editing and conditional formatting. It uses Anthropic's most intelligent model for complex spreadsheet tasks.
Claude Haiku 4.5 Released
Anthropic releases Claude Haiku 4.5, its fastest and most efficient model with near-frontier performance. Designed for real-time applications, high-volume processing, and cost-sensitive deployments requiring strong reasoning. Available across the API, Amazon Bedrock, and Vertex AI.
Claude Memory Feature Launched
Anthropic announces persistent memory capabilities for Claude, allowing it to retain user and team projects, preferences, and context across conversations. Memory is rolling out to Team and Enterprise plan users. A new Incognito chat mode is simultaneously introduced for all Claude users who prefer no memory retention.
Anthropic Report: Enterprises Using AI for Task Delegation
Anthropic releases a report finding that businesses primarily use AI for automation rather than collaboration, with three-quarters of companies using Claude for 'full task delegation.' The report analyzes real enterprise usage patterns, informing how Anthropic positions future product features.
Anthropic Restricts Sales to Adversarial Nation Entities
Anthropic announces it will stop selling its products to groups majority-owned by Chinese, Russian, Iranian, or North Korean entities due to national security concerns. This policy restricts access to Claude for potential adversarial uses while maintaining broad global commercial availability.
Series F Funding — $13B at $183B Valuation
Anthropic closes a $13 billion Series F funding round co-led by ICONIQ Capital, Fidelity Management & Research, and Lightspeed Venture Partners, with participation from the Qatar Investment Authority and others. Post-money valuation reaches $183 billion, making Anthropic one of the most valuable private technology companies globally.
Claude Sonnet 4.5 Released
Anthropic releases Claude Sonnet 4.5, achieving a landmark 77.2% on SWE-bench Verified — the highest score ever at launch. The model features 30+ hour focus windows for long-running agent tasks and 0% error rate on Replit coding benchmarks. It is positioned as the world's best coding model at launch.
Anthropic Reaches $5B Annualized Revenue
Anthropic's annualized revenue reaches approximately $5 billion by August 2025, up from ~$1 billion at the start of 2025 — a 5x increase in 8 months. Claude Code is a key driver, reaching a $1 billion ARR run rate within months of its May 2025 GA launch. The company has 300,000+ business customers by August 2025.
Claude Opus 4.1 Released
Anthropic releases Claude Opus 4.1, an updated version of its flagship Opus model with improvements in multi-step reasoning and agentic tasks. Anthropic also enables a capability for Opus 4 and 4.1 to end conversations that remain persistently harmful or abusive as a last resort safety measure.
Claude in Chrome Extension Launched
Anthropic launches Claude in Chrome as a browser extension, enabling Claude to assist with web browsing tasks directly in the Chrome browser. This enables AI-assisted browsing, content summarization, and research directly within the browser without switching to a separate application.
Claude Sonnet 4 1M Token Context Window
Anthropic expands Claude Sonnet 4's context window to 1 million tokens (5x increase), enabling processing of entire large codebases (75,000+ lines), dozens of research papers, or extensive document sets in a single request. Long context is initially available in public beta on the API, Amazon Bedrock, and Vertex AI.
US Department of Defense $200M AI Contract
The US Department of Defense awards Anthropic a $200 million contract for AI in the military, alongside contracts to Google, OpenAI, and xAI. This represents a major government revenue milestone for Anthropic and validates its enterprise-grade AI for critical government applications.
Artifacts Upgraded with App-Building & Hosting
Anthropic upgrades the Artifacts feature to allow users to build, host, and share interactive AI-powered apps directly from Claude.ai. Users describe app ideas in plain language and Claude generates functional tools like data analyzers and study aids. Usage costs are attributed to the viewer's subscription, not the creator's.
Claude Gov Model Announced
Anthropic announces 'Claude Gov,' a specialized version of Claude for US government and national security use cases. By June 2025, the model is reportedly in use at multiple US national security agencies, signaling Anthropic's deeper entry into the federal and defense markets.
Anthropic Inaugural Developer Conference
Anthropic hosts its first developer conference, showcasing Claude 4 capabilities, Claude Code GA, and the new API features for agent builders. The conference marks Anthropic's growing developer community engagement and commitment to the developer ecosystem.
Interleaved Thinking (Beta) Launched
Anthropic releases interleaved thinking in public beta, enabling Claude to think between tool calls during agentic workflows. This allows the model to reason about intermediate results before deciding on next steps, significantly improving performance on complex multi-step tasks. Enabled via anthropic-beta header.
Files API, Code Execution Tool & MCP Connector in API
Anthropic releases four new API capabilities alongside Claude 4: the Files API for persistent file storage across sessions, a sandboxed Code Execution tool for running Python safely, an MCP Connector for connecting to remote MCP servers without client code, and extended 1-hour prompt caching. These enable developers to build significantly more powerful AI agents.
Claude Code Generally Available
Anthropic announces Claude Code is now generally available, with new VS Code and JetBrains IDE extensions for inline code editing. The Claude Code SDK is released for building custom agents. Claude Code also gains GitHub Actions support and passes $1 billion ARR run rate within months of launch.
Claude 4: Opus 4 & Sonnet 4 Launched
Anthropic releases Claude 4, introducing Claude Opus 4 (most powerful, 72.5% SWE-bench Verified, 43.2% Terminal-bench) and Claude Sonnet 4 (72.7% SWE-bench). Both are hybrid models with standard and extended thinking modes. Anthropic classifies Opus 4 as an ASL-3 safety level model. Pricing: Opus 4 at $15/$75, Sonnet 4 at $3/$15 per million tokens.
Web Search API Tool Launched
Anthropic launches a web search tool via the API, enabling Claude to access real-time information from the internet during API calls. This enables developers to build applications where Claude can search the web for current information, overcoming the model's training data cutoff.
Claude Max Plan Launched
Anthropic introduces the Claude Max plan, offering the highest usage limits and priority access to new features for individual users. The Max plan is positioned above Claude Pro for power users and professionals who require intensive AI usage. Includes access to both standard and extended thinking modes.
Research: Multilingual Processing & Planning Ahead in LLMs
Anthropic researchers publish findings that multilingual LLMs partially process information in a conceptual space before converting to the target language. The research also finds evidence that LLMs plan ahead — for example, when writing poetry, Claude identifies potential rhyming words before generating lines. These findings have implications for interpretability and alignment.
Anthropic & Databricks Integration Partnership
Databricks and Anthropic announce that Claude will be integrated into the Databricks Data Intelligence Platform, enabling AI-powered analytics and data intelligence at scale for enterprise data teams using the Databricks ecosystem.
Token-Efficient Tools & Cache-Aware Rate Limits
Anthropic releases several API improvements for Claude 3.7 Sonnet: cache-aware ITPM rate limits (cached tokens don't count against input limits), simplified automatic prompt caching, and token-efficient tool use reducing output token consumption by up to 70%. These changes significantly improve throughput and reduce costs for agent workflows.
Series E Funding — $3.5B at $61.5B Valuation
Anthropic closes a $3.5 billion Series E funding round led by Lightspeed Venture Partners with participation from Fidelity, Cisco, Salesforce, D1 Capital, and others. The post-money valuation reaches $61.5 billion, making Anthropic the second most valuable AI startup after OpenAI.
Google Invests Additional $1B in Anthropic
Google agrees to invest an additional $1 billion in Anthropic, strengthening the existing strategic partnership. This brings Google's total committed investment to approximately $3 billion, maintaining Google as a key investor alongside Amazon.
Web Search Feature Launched for Paying Users
Anthropic adds a web search feature to Claude, enabling the model to access real-time information from the internet. Initially available to paying users in the United States. Free users gained access in May 2025. This bridges Claude's knowledge cutoff gap and enables current event awareness.
Claude 3.5 Haiku Gains Image Input Support
Anthropic adds image input capabilities to Claude 3.5 Haiku, making it the last remaining modern Claude model to gain vision support. This completes vision capabilities across all active Claude 3.5 models, enabling image analysis in cost-effective and high-volume deployments.
Claude Code Research Preview Launched
Anthropic introduces Claude Code, a command-line interface (CLI) tool for agentic coding, as a limited research preview. Claude Code can search and read code, edit files, write and run tests, commit and push to GitHub, and use command-line tools. Developers can delegate substantial engineering tasks to Claude directly from their terminal.
Claude 3.7 Sonnet & Extended Thinking Launched
Anthropic releases Claude 3.7 Sonnet, the company's first hybrid reasoning model, capable of both near-instant responses and extended step-by-step thinking. API users can control the thinking budget (token limit) for fine-grained speed/quality tradeoffs. Claude 3.7 Sonnet achieves 62.3% on SWE-bench Verified (70.3% with scaffolding) and is immediately available on all Claude plans, API, Bedrock, and Vertex AI at the same price as Claude 3.5 Sonnet.
UK DSIT Memorandum of Understanding Signed
Anthropic signs a Memorandum of Understanding with the UK Department for Science, Innovation and Technology (DSIT), establishing a framework for AI collaboration with the UK government. This later leads to the GOV.UK AI assistant pilot announced in January 2026.
Anthropic Economic Index Launched
Anthropic introduces the Anthropic Economic Index, a continuous research series analyzing the impact of AI on labor markets and the broader economy. The initial report analyzes millions of anonymized Claude conversations to understand which job tasks AI is affecting most and how, providing data-driven insights into AI's economic impact.
Computer Use Tool Updated (January 2025)
Anthropic updates the Computer Use tool (computer_20250124) with new command options including hold_key, left_mouse_down, left_mouse_up, scroll, triple_click, and wait. These additions expand Claude's ability to interact with complex GUI applications and improve reliability of computer use workflows.
Anthropic Reaches 1,000+ Employees
Anthropic's employee count surpasses 1,097, representing a 471% increase since 2022. CEO Dario Amodei notes intentional slowing of hiring to maintain quality, focusing on senior hires from top companies and theoretical physicists. The company also hires its first Chief Commercial Officer, Paul Smith, and first Managing Director of International.
Citations API Feature Launched
Anthropic launches a Citations capability in the API, allowing Claude to provide source attribution for information in its responses. Citations enable developers to build applications where Claude can reference specific documents or passages, improving transparency and verifiability of AI-generated content.
Prompt Caching Generally Available
Anthropic announces prompt caching is now generally available on the Anthropic API, graduating from public beta. Prompt caching is also made available in preview on Amazon Bedrock and Google Cloud's Vertex AI. Developers can now reliably cache up to 5 minutes of context across API calls.
Model Context Protocol (MCP) Announced & Open-Sourced
Anthropic open-sources the Model Context Protocol (MCP), a new universal standard for connecting AI assistants to external data sources and tools. MCP solves the 'M×N integration problem' by providing a single protocol instead of custom connectors for each data source. Initial reference servers included Google Drive, Slack, GitHub, Postgres, and Puppeteer. Created by Anthropic engineers David Soria Parra and Justin Spahr-Summers.
Responsible Scaling Policy Update (v1.1)
Anthropic publishes an updated version of its Responsible Scaling Policy, refining the AI Safety Level definitions and evaluation procedures. The update reflects learnings from deploying frontier models and incorporates feedback from the AI safety community.
Palantir & AWS Partnership for US Defense
Anthropic partners with Palantir and Amazon Web Services to provide Claude models to US intelligence and defense agencies. The partnership enables classified AI deployments, marking Anthropic's entry into the government and defense sector.
Amazon Invests Additional $4B — Total $8B
Amazon announces an additional $4 billion investment in Anthropic in November 2024, doubling its total commitment to $8 billion. This makes Amazon by far the largest investor in Anthropic and solidifies AWS as the preferred cloud and chip provider for training and deployment.
Computer Use Feature (Beta) Launched
Anthropic introduces Computer Use in public beta — enabling Claude to interact with computers by interpreting screenshots and simulating keyboard and mouse inputs. Claude can navigate GUIs, fill forms, write and run code, and perform web browsing autonomously. This represents a major step toward fully agentic AI systems.
Claude 3.5 Haiku Released
Anthropic releases Claude 3.5 Haiku, its fastest and most affordable Claude 3.5 model. Claude 3.5 Haiku surpasses Claude 3 Opus on several benchmarks while running at lower cost, making high-quality AI accessible for real-time and high-volume applications. A price increase was announced alongside this release.
Claude 3.5 Sonnet v2 (Updated) Released
Anthropic releases an upgraded version of Claude 3.5 Sonnet with improved coding, reasoning, and instruction following. This version introduced as claude-3-5-sonnet-20241022 on the API. Alongside this release, Anthropic also released claude 3.5 Haiku, making faster models more capable.
Dario Amodei Publishes 'Machines of Loving Grace' Essay
Anthropic CEO Dario Amodei publishes 'Machines of Loving Grace,' a landmark essay proposing that AI-enabled biology and medicine could compress 50-100 years of human biological progress into 5-10 years. The essay articulates a positive vision for transformative AI and becomes widely cited in discussions about AI's potential impact.
Contextual Retrieval Technique Published
Anthropic introduces Contextual Retrieval, a technique that dramatically improves RAG (Retrieval-Augmented Generation) systems by adding context to each chunk of a knowledge base before embedding. When combined with BM25 retrieval, it reduces failed retrievals by up to 67%. Made cost-effective by Anthropic's prompt caching feature.
Message Batches API Launched
Anthropic launches the Message Batches API, allowing developers to submit large batches of requests for asynchronous processing at a 50% discount compared to standard API pricing. This enables cost-effective processing of bulk workloads such as data pipelines, evaluations, and content generation at scale.
Claude Enterprise Plan Launched
Anthropic launches Claude Enterprise, its premium business plan with a 500,000-token context window (2.5x the Team plan), GitHub integration, enhanced admin controls, and SSO. Early adopters include GitLab, Midjourney, and Menlo Ventures. Claude Enterprise comes with Projects and Artifacts. Custom pricing based on organizational size.
Prompt Caching Public Beta
Anthropic launches prompt caching in public beta on the Anthropic API, allowing developers to cache frequently used context between API calls. Caching reduces costs by up to 90% and latency by up to 85% for long prompts. Available for Claude 3.5 Sonnet, Claude 3 Opus, and Claude 3 Haiku.
Projects Feature Launched on Claude.ai
Anthropic launches Projects, a new collaboration feature on Claude.ai allowing teams to create self-contained workspaces for specific initiatives. Users can upload files and background context to customize Claude's behavior within a project. Projects enable persistent memory of conversations and documents within a shared team context.
Prompt Evaluation Tools in Developer Console
Anthropic launches new evaluation features in the Developer Console, including automatic test case generation and side-by-side output comparison. Developers can now generate, test, and evaluate prompts directly in the Anthropic Console using Claude itself as the evaluation engine.
Claude Android App Launched
Anthropic launches the Claude Android app, extending mobile access to Android users following the May 2024 iOS launch. The Android app provides the same core features including chat history sync, vision capabilities, and access to all Claude models.
Claude 3.5 Sonnet Achieves 64% on Agentic Coding Eval
Claude 3.5 Sonnet sets new state-of-the-art results on Anthropic's internal agentic coding evaluation, solving 64% of problems compared to Claude 3 Opus's 38%. The model can independently write, edit, and execute code with sophisticated troubleshooting, marking a major step toward autonomous software development.
Tool Use Generally Available on Anthropic API
Anthropic announces tool use (function calling) is now generally available across the Claude API, Amazon Bedrock, and Google Cloud's Vertex AI. This allows developers to give Claude access to external APIs, databases, and custom functions, enabling powerful agentic workflows.
Claude 3.5 Sonnet Released + Artifacts Feature
Anthropic releases Claude 3.5 Sonnet, outperforming even Claude 3 Opus on most benchmarks while running at Sonnet speed and pricing. The model achieves 64% on agentic coding tasks. Alongside the model, Anthropic launches Artifacts — a new feature creating a side-by-side workspace for code, documents, and web designs, enabling real-time preview and editing of AI-generated content.
Mechanistic Interpretability: Golden Gate Claude
Anthropic publishes a mechanistic interpretability paper identifying millions of 'features' (internal concept representations) in Claude 3 Sonnet using dictionary learning. As a demonstration, Anthropic releases 'Golden Gate Claude' — a version with the Golden Gate Bridge feature strongly activated, making it obsessively reference the bridge. This is a landmark achievement in understanding AI internals.
Claude's Character and Values (Model Spec) Published
Anthropic publishes a comprehensive 'Model Spec' document articulating Claude's core character, values, and approach to being helpful, honest, and harmless. The document describes Claude as having genuine intellectual curiosity, warmth, and a commitment to honesty, going beyond simple rule-following to character-based AI alignment.
Key Leadership Hires: Jan Leike, Krishna Rao, Mike Krieger
Anthropic makes several high-profile executive hires in May 2024: Jan Leike (former OpenAI safety leader) to lead a new safety team, Krishna Rao as Anthropic's first CFO, and Instagram co-founder Mike Krieger as Head of Product. These hires signal Anthropic's transition from research-focused startup to mature AI company.
Claude Team Plan & iOS App Launched
Anthropic launches the Claude Team plan ($30/user/month, minimum 5 users) for business collaboration, alongside the first Claude iOS app for iPhone and iPad. The iOS app syncs with web chats and supports Claude 3's vision capabilities for image analysis. Team plan provides higher usage limits, admin controls, and access to all Claude 3 models.
Amazon Completes $4B Total Investment
Amazon completes its $4 billion investment in Anthropic with an additional $2.75 billion payment, fulfilling the commitment made in September 2023. AWS becomes the primary cloud provider for Anthropic and the first global cloud to host Claude 3 models. This cements AWS as a key distribution partner for enterprise deployments.
Claude 3 Available on Google Cloud Vertex AI
Claude 3 models (Haiku, Sonnet, Opus) become available on Google Cloud's Vertex AI platform, expanding enterprise access beyond the Anthropic API. This marks a major distribution milestone for Claude in enterprise cloud environments.
Claude 3 Haiku Released
Anthropic makes Claude 3 Haiku generally available, the fastest and most affordable model in the Claude 3 family. Haiku is designed for near-instant responsiveness and high-volume use cases. It follows Opus and Sonnet which launched simultaneously on March 4, 2024, with Haiku completing the three-tier lineup shortly after.
Claude 3 Available on Amazon Bedrock
Claude 3 models (Haiku, Sonnet, Opus) become available on Amazon Bedrock, with AWS being the first global cloud to host Claude 3 models. This aligns with Amazon's $4 billion investment and primary cloud provider arrangement, enabling enterprise customers to access Claude's latest capabilities through AWS infrastructure.
Claude Vision — Image Upload and Analysis
With the Claude 3 launch, Anthropic introduces multimodal capabilities enabling Claude to process and analyze images alongside text. Users can upload photos, documents, diagrams, and screenshots for analysis. Claude 3 Opus and Sonnet become Anthropic's strongest vision models, surpassing competitors on standard vision benchmarks.
Anthropic Releases Claude Safety Evaluation Benchmarks
Anthropic publishes its safety evaluation methodology and benchmarks for assessing Claude's performance on harmful content generation, deception resistance, and dangerous capability evaluations. These evaluations are used to assess models against RSP thresholds before deployment.
Claude 3 Family Launched — Haiku, Sonnet, Opus
Anthropic releases the Claude 3 model family with three tiers: Claude 3 Haiku (fast, affordable), Claude 3 Sonnet (balanced), and Claude 3 Opus (most capable). Claude 3 introduces multimodal capabilities, accepting both text and images. Claude 3 Opus sets new industry benchmarks across reasoning, coding, and multilingual tasks, outperforming GPT-4 on multiple evaluations.
Series D Funding — $750M (Menlo Ventures)
Menlo Ventures leads a $750 million Series D funding round for Anthropic. The round increases Anthropic's valuation significantly and provides additional runway for model development, safety research, and commercial expansion.
Claude's Constitution (Model Spec) Published Publicly
Anthropic publicly releases 'Claude's Constitution' — the set of principles and values used to guide Claude's behavior through Constitutional AI training. Publishing the constitution increases transparency about how Claude is trained and what ethical guidelines it follows, inviting industry-wide adoption of similar approaches.
Claude 2.1 Released — 200K Context Window
Anthropic releases Claude 2.1, doubling the context window to 200,000 tokens (approximately 500 pages of written content). Claude 2.1 reduces hallucinations by approximately 2x versus Claude 2.0 and adds beta support for tool use including API calls and web search integration. This targets enterprise use cases requiring analysis of large documents.
Claude on Amazon Bedrock — General Availability
Claude models become generally available on Amazon Bedrock, Amazon's managed AI service, following the partnership announcement. This enables AWS customers to access Claude through the Bedrock API with enterprise security and compliance features, significantly expanding Claude's distribution.
Google Invests $500M (Committed Up to $2B)
Google invests $500 million in Anthropic and commits to invest up to $2 billion total over time, deepening the existing cloud partnership. This makes Google one of Anthropic's largest strategic investors alongside Amazon. Claude models are made available on Google Cloud's Vertex AI platform.
Amazon Investment Announced — Up to $4B
Amazon announces an investment of up to $4 billion in Anthropic, starting with an initial $1.25 billion. As part of the deal, Anthropic will use AWS as its primary cloud provider and make Claude models available on Amazon Bedrock. Amazon becomes a minority stakeholder with no board seat.
Responsible Scaling Policy (RSP) v1 Published
Anthropic publishes its first Responsible Scaling Policy, committing to AI Safety Levels (ASLs) and self-imposed thresholds for when to pause or adjust model development based on capability evaluations. This makes Anthropic one of the first major AI labs to adopt a formal, public safety commitment tied to model capabilities.
Research Paper: Frontier Threats Red Teaming for AI Safety
Anthropic publishes research on frontier threat red teaming methodologies for AI safety evaluation. The paper addresses how to evaluate potential harms from frontier AI models, contributing to the field of AI safety evaluation practices and informing Anthropic's Responsible Scaling Policy.
Frontier AI Safety Commitment Signed
Anthropic is among the first AI companies to sign voluntary commitments with the US White House and other global bodies on frontier AI safety. These commitments include sharing safety evaluations, red-teaming results, and investing in safety research.
Claude Pro Subscription Plan Launched
Anthropic launches the Claude Pro subscription plan at $20/month, providing paying users with significantly higher rate limits, priority access during high-traffic periods, and early access to new features. Claude Pro launches simultaneously with Claude 2 and Claude.ai.
Claude 2.0 & Claude.ai Web App Launch
Anthropic releases Claude 2, a major model upgrade with improved performance on coding, reasoning, and safety. Alongside the model, Anthropic launches Claude.ai — its consumer-facing web application — making Claude available to the general public without API access. Claude 2 becomes the first publicly available model. Claude Pro subscription at $20/month also launches.
Policy Paper: Charting a Path to AI Accountability
Anthropic publishes a policy paper outlining its views on AI accountability mechanisms and regulatory approaches. The paper advocates for third-party audits, safety evaluations, and industry-wide standards for responsible AI development.
Zoom and Salesforce Investments & Partnerships
Zoom and Salesforce invest in Anthropic as part of the Series C round. Both companies sign partnership agreements to integrate Claude into their respective platforms. This marks Anthropic's first major enterprise software platform integrations.
Series C Funding — $450M
Anthropic raises $450 million in a Series C round led by Spark Capital, with participation from Google, Salesforce Ventures, Sound Ventures, Zoom Ventures, and others. Yasmin Razavi from Spark Capital joins Anthropic's Board of Directors. Funds support continued work on helpful, harmless, and honest AI systems.
100K Token Context Window Announced
Anthropic announces Claude's ability to process up to 100,000 tokens in a single context window — equivalent to roughly 75,000 words or an entire book. This was a major step change in context length, enabling Claude to analyze entire codebases, lengthy legal documents, or long narratives in one pass.
Claude 1.3 Released
Anthropic releases Claude 1.3, the first major upgrade to Claude 1.0. The update improves Claude's reasoning, reliability, and instruction following. This version represents Anthropic's first iteration on its deployed model based on real-world usage feedback.
Claude 1.0 Public Launch
Anthropic publicly launches Claude 1.0, its first commercially available AI assistant, following a limited beta period. Claude is positioned as a helpful, harmless, and honest AI assistant based on Constitutional AI training. The model is made available via API and through partner integrations including Slack.
Scale AI Partnership
Anthropic announces a partnership with Scale AI to support data labeling, evaluation, and model training data pipelines. Scale AI's annotation infrastructure helps Anthropic generate high-quality training data for Claude models.
Anthropic Publishes 'Core Views on AI Safety'
Anthropic publishes a document articulating its core views on AI safety, explaining why it believes AI may be among the most transformative and dangerous technologies ever developed, and why it has chosen to proceed anyway through a safety-focused approach. This document establishes Anthropic's public safety philosophy.
Claude Public Waiting List Opens
Following the Google Cloud partnership announcement, Anthropic opens a public waiting list for early access to Claude. Claude had been available as a Slack integration during its closed beta. This marks the beginning of public deployment of Claude's AI assistant capabilities.
Google Cloud Partnership & $300M Investment
Anthropic announces Google Cloud as its preferred cloud provider, while Google invests approximately $300 million for roughly a 10% stake. Claude and Anthropic AI systems will run on Google Cloud going forward. The partnership enables Anthropic to leverage Google Cloud's TPUs and GPUs to train and deploy its models.
Long-Term Benefit Trust Governance Structure
Anthropic establishes the Long-Term Benefit Trust (LTBT), an internal governance body tasked with overseeing the company's public benefit mission. The trust maintains oversight independent of investor pressure, ensuring Anthropic can prioritize safety even if it conflicts with short-term commercial interests.
Constitutional AI Paper Published
Anthropic releases its landmark Constitutional AI (CAI) paper, introducing a novel approach to AI alignment where a list of principles ('constitution') guides AI behavior without extensive human feedback. The technique allows AI systems to self-critique using predefined ethical guidelines, establishing a new paradigm for training safe and helpful AI.
First Claude Model Trained (Internal)
Anthropic completes training of the first version of Claude internally but does not release it publicly. The decision not to release cites the need for further safety testing and a desire to avoid initiating a potentially harmful AI race. This model lays the groundwork for the eventual Claude 1 public release.
Series B Funding — $580M
Anthropic raises $580 million in a Series B round led by Sam Bankman-Fried of FTX, with $500 million from Alameda Research. Other investors include Caroline Ellison, Jaan Tallinn, and others. The funding accelerates development of large-scale models and AI safety research teams. The FTX connection later became controversial after FTX's collapse in November 2022.
Research Paper: Training a Helpful and Harmless Assistant with RLHF
Anthropic introduces preference modeling and reinforcement learning from human feedback (RLHF) methodology for training AI assistants to be 'helpful and harmless.' The paper was the first to explore online RLHF training and measure the tension between helpfulness and harmlessness, becoming a foundational safety alignment technique.
Research Paper: Predictability and Surprise in Large Generative Models
Anthropic publishes one of its first research papers investigating emergent capabilities in large language models. The paper finds that while model accuracy improves consistently with scale, certain abilities (e.g., three-digit addition) appear suddenly at certain parameter thresholds. This raises concerns about unpredictability of future AI capabilities.
Series A Funding — $124M
Anthropic raises $124 million in its Series A funding round at a $550 million valuation. The round was led by Jaan Tallinn (co-founder of Skype) with participation from Dustin Moskovitz, Eric Schmidt, the Center for Emerging Risk Research, and others. Funds support research on large-scale AI safety and prototype model development.
Anthropic Founded
Dario Amodei, Daniela Amodei, and seven colleagues leave OpenAI to found Anthropic as a public benefit corporation (PBC) in San Francisco. The company was initially called 'AI Safety Lab' before being renamed. Their mission: responsible development of AI for long-term human benefit.
Based on publicly available Anthropic announcements. Dates reflect original release dates. Curated by claudeai.directory.