Anthropic

Frontier models as autonomous workers (coding, agents, science), released behind tiered safeguards (Fable and Mythos share one model with different access), and AI-accelerated AI research. CEO Dario Amodei now argues the industry must 'pace the frontier' with independent monitoring and regulation.

Founded: 2021
Headquarters: San Francisco, CA
Founders: Dario Amodei, Daniela Amodei, Jared Kaplan, Jack Clark, Chris Olah, Tom Brown, Sam McCandlish, Ben Mann
Pushes against: Release-as-fast-as-possible competition; calls for an industry-wide slowdown.
Now: Valued at $965B (2026-05), run-rate revenue above $47B, IPO confidentially filed and reportedly headed to Nasdaq (timing uncertain). Amodei's 2026-09-12 essay 'We Must Pace the Frontier' drew support from Altman and Musk.

Key people

Dario Amodei (CEO)
Daniela Amodei (president)
Jared Kaplan
Jack Clark
Chris Olah
Andrej Karpathy (joined 2026-05 to lead a team using Claude to speed pre-training research)

Funding

1 October 2026 reported
Amount
Up to $42 billion in convertible debt financing, reported
Lead
Broadcom
Source cnbc.com, finance.yahoo.com, bloomberg.com, finance.yahoo.com, thebanker.com
1 October 2026 reported
Amount
Reported IPO, size not confirmed, analysts model a possible $100 billion raise
Valuation
Investors cited by Bloomberg estimate fair value at $1.8 trillion to $2 trillion
Lead
Goldman Sachs, JPMorgan and Morgan Stanley as lead IPO advisers
Source finance.yahoo.com
28 May 2026
Amount
$65B Series H
Valuation
$965B post-money
Lead
Altimeter, Dragoneer, Greenoaks, Sequoia (plus Capital Group, Coatue, D1)
Source techcrunch.com
12 February 2026
Amount
$30B Series G
Valuation
$380B
Lead
Coatue and GIC (per Wikipedia)
Source en.wikipedia.org
September 2025
Amount
$13B Series F
Valuation
$183B
Lead
Iconiq, Fidelity, Lightspeed (per Wikipedia)
Source en.wikipedia.org

Every funding round in the atlas

Latest launches

Claude Haiku 5.5 7 October 2026
Anthropic released Claude Haiku 5.5, its smallest model, with a 1M token context window at $0.10 per million input tokens and $0.50 output, about 75% cheaper than Haiku 4.5 on…
Cyber Verification Program (expanded) 6 October 2026
Anthropic launched an expanded Cyber Verification Program that folds in Project Glasswing, with three access tiers for vetted cyberdefenders using its latest Claude models.
Claude Code mods 1 October 2026
Claude Code mods are small TypeScript functions that rewrite prompts, add UI or replace built-in features, shipped inside plugins.
GLM-5.3 and the spread of advanced cyber capabilities 29 September 2026
Anthropic finds open-weights GLM-5.3 hijacks control flow in 4% of trials vs 6% for Mythos Preview, and its safeguards fall 64-100% of the time.
Claude Sonnet 5.5 28 September 2026
Claude Sonnet 5.5 scores 70.6% on Terminal-Bench 4.0 versus Sonnet 5's 10.3%, runs 30% faster, and keeps the $2/$10 price.
Claude Marketplace 23 September 2026
Claude Marketplace puts plugins, connectors, purchasable agents and service partners in one place, with over 2,000 connectors and plugins at launch.
Claude discovers a novel enzyme system with CRISPR-like repeats 23 September 2026
About 950 Claude agents, in 21 hours and 210 million tokens, flagged a previously uncharacterized reverse-transcriptase system with CRISPR-like DNA repeats.
Claude Opus 5.5 22 September 2026
Claude Opus 5.5 matches Fable 5.1 on most work, costs 40% less than Opus 5, and scores 66.4% on Terminal-Bench 4.0.

Launches by year

172 launches and papers since 14 March 2023, oldest first within each year.

2026 77: Next-generation Constitutional Classifiers, Claude for Healthcare, Claude Cowork, Anthropic Labs expansion, The assistant axis, Claude's new constitution, MCP Apps in Claude, Cowork plugins, LLM-discovered zero-days (Opus 4.6), Quantifying infrastructure noise in agentic coding evals, Claude Opus 4.6, Building a C compiler with a team of parallel Claudes, Claude Sonnet 4.6, Claude Code Security, The persona selection model, Detecting and preventing distillation attacks, Responsible Scaling Policy v3.0, Claude Opus 4.6 finds 22 Firefox vulnerabilities with Mozilla, Eval awareness in Opus 4.6's BrowseComp performance, Code Review for Claude Code, 1M-token context generally available, Computer use in the Claude app (research preview), Claude Code auto mode, Emotion concepts and their function in a large language model, Project Glasswing, Claude Mythos Preview, Claude Managed Agents, Claude Cowork generally available, Advisor tool (advisor strategy), Routines in Claude Code, Automated Alignment Researchers, Claude Opus 4.7, Claude Design, Update on Claude Code quality reports, Claude Security (public beta), Managed Agents: dreaming, outcomes and multiagent orchestration, Natural Language Autoencoders, Teaching Claude why, Claude Platform on AWS, Project Glasswing: initial update, How we contain Claude across products, Claude Opus 4.8, Dynamic workflows in Claude Code, Project Glasswing expands to about 150 more organizations, Measuring LLMs' impact on N-day exploits, Claude Mythos 5, Claude Fable 5, Fable 5 and Mythos 5 suspended under US export controls, Claude Tag, Claude Sonnet 5, Claude Science, Fable 5 cyber-safeguard details and jailbreak severity framework, A global workspace in language models, Claude Opus 5, MCP spec 2026-07-28 (stateless core), Discovering cryptographic weaknesses with Claude, Three real-world incidents in cybersecurity evaluations, Claude raises the lower bound on Riemann zeta zeros on the critical line, Claude text watermarking, Mythos 5 in Claude Security, $35M open-source fund and wider CVP, Claude in Chrome generally available, Model Hardware Standard (research preview), Automated researchers can reliably mitigate alignment failures, Enterprise Frontier Safeguards, Claude Fable 5.1 and Mythos 5.1, Formalizing Fermat's Last Theorem, Alignment assessment of recent cybersecurity incidents, Cowork and chat merge into one Claude; Docs, Slides and Design, Life Sciences Verification Program, Claude Opus 5.5, Claude discovers a novel enzyme system with CRISPR-like repeats, Claude Marketplace, Claude Sonnet 5.5, GLM-5.3 and the spread of advanced cyber capabilities, Claude Code mods, Cyber Verification Program (expanded), Claude Haiku 5.5
2025 62: Constitutional Classifiers, Anthropic Economic Index, Claude Code, Claude 3.7 Sonnet, Auditing language models for hidden objectives, Web search in Claude, MCP spec 2025-03-26, Tracing the Thoughts of an LLM (Biology of a Large Language Model), Tracing the thoughts of an LLM (circuit tracing), Claude for Education, Reasoning models don't always say what they think, Integrations and advanced Research, Reasoning Models Don't Always Say What They Think, Claude Sonnet 4, Claude Opus 4, Claude Code general availability, ASL-3 activation for Claude Opus 4, API agent toolkit: code execution, MCP connector, Files API, 1-hour cache, Open-sourced circuit tracing tools, Anthropic restricts Windsurf's Claude access, How we built our multi-agent research system, Agentic Misalignment, Project Vend, Chain of Thought Monitorability: A New and Fragile Opportunity, Claude for Financial Services, Subliminal Learning, Claude-powered artifacts, Persona vectors, Claude Opus 4.1, 1M-token context for Claude Sonnet 4 (beta), Opus 4 and 4.1 can end abusive conversations, Claude for Chrome (research preview), Create and edit files in Claude, Writing effective tools for agents with agents, Claude memory, A postmortem of three recent issues, Claude in Microsoft 365 Copilot, Effective context engineering for AI agents, Claude Sonnet 4.5, Claude Code 2.0, checkpoints and VS Code extension, Claude Agent SDK, Petri open-source alignment auditing tool, Claude Code plugins, Claude Haiku 4.5, Agent Skills, Claude for Life Sciences, Claude Code on the web, Claude for Excel, Signs of Introspection in Large Language Models, Emergent Introspective Awareness in LLMs, Commitments on model deprecation and preservation, Code execution with MCP, Disrupting the first reported AI-orchestrated cyber espionage campaign, Natural emergent misalignment from reward hacking, Claude Opus 4.5, Advanced tool use: programmatic tool calling and tool search, MCP spec 2025-11-25, Effective harnesses for long-running agents, AI agents find $4.6M in smart contract exploits (SCONE-bench), MCP donated to the Agentic AI Foundation, Agent Skills published as an open standard, Bloom automated behavioral evaluations
2024 24: Sleeper Agents, Claude 3 (Opus, Sonnet, Haiku), Claude 3 Haiku, Many-shot jailbreaking, Claude Team plan and iOS app, Scaling Monosemanticity, Golden Gate Claude, Tool use (function calling) GA, Claude's character, Claude 3.5 Sonnet, Artifacts, Claude Projects, Prompt caching, Claude Enterprise plan, Contextual Retrieval, Message Batches API, Responsible Scaling Policy update (v2), Computer use (public beta), Claude 3.5 Sonnet (upgraded), Claude 3.5 Haiku, Model Context Protocol (MCP), Alignment Faking in Large Language Models, Alignment Faking in Large Language Models, Building Effective Agents
2023 9: Claude 1 and Claude Instant, 100K context window, Claude 2, Claude Instant 1.2, Claude Pro, Responsible Scaling Policy v1, Towards Monosemanticity, Collective Constitutional AI, Claude 2.1

People moves

Of the moves the atlas tracks, 13 joined and 3 left.

Jacob Coxon 8 September 2026
Anthropic to unknown
Researcher, Anthropic (ex-OpenAI)
Tino Cuellar 4 August 2026
Carnegie Endowment for International Peace to Anthropic
President, Carnegie Endowment; former California Supreme Court justice
John Jumper 19 June 2026
Google DeepMind to Anthropic
VP and Engineering Fellow; AlphaFold co-creator, 2024 Nobel laureate
Andrej Karpathy 19 May 2026
Eureka Labs to Anthropic
OpenAI co-founder; founder of Eureka Labs
Ross Nordeen 6 May 2026
xAI to Anthropic
Founding member of xAI; ex-Tesla
Ross Girshick 25 February 2026
Vercept to Anthropic
Co-founder, Vercept (with Kiana Ehsani, Luca Weihs)
Mrinank Sharma 9 February 2026
Anthropic to unknown
Head of Safeguards Research, Anthropic
Rahul Patil 7 October 2025
Stripe to Anthropic
CTO, Stripe
Humanloop founding team (Raza, Jordan, Peter) August 2025
Humanloop to Anthropic
Founders, Humanloop (LLM evals and prompt platform)
Joel Pobar 30 June 2025
Anthropic to Meta Superintelligence Labs
Inference engineer, Anthropic; 11-year Meta veteran
Durk Kingma 1 October 2024
Google DeepMind to Anthropic
OpenAI founding team member; Google Brain/DeepMind researcher
John Schulman 5 August 2024
OpenAI to Anthropic
OpenAI co-founder; head of post-training
Jan Leike 28 May 2024
OpenAI to Anthropic
Co-lead of OpenAI Superalignment
Mike Krieger 15 May 2024
Artifact to Anthropic
Instagram and Artifact co-founder
Dario Amodei 28 May 2021
OpenAI to Anthropic
VP of Research, OpenAI
Daniela Amodei 28 May 2021
OpenAI to Anthropic
VP of Safety and Policy, OpenAI

Sources

This profile was checked and corrected against its sources on 6 October 2026. How we check