Anthropic

Anthropic is an American artificial intelligence safety company founded in 2021 and headquartered in San Francisco. It develops the Claude family of large-language models and related products including Claude Code, Claude Cowork, and the Anthropic API. As of April 2026, it is among the leading frontier AI labs by capability, valuation, and enterprise adoption.

At a glance

• Founded: 2021 in San Francisco
• Status: Private. Public benefit corporation (PBC).
• Funding: Approximately $73 billion cumulative. Most recent documented round: $30 billion Series G at a $380 billion post-money valuation in February 2026. Reports in April 2026 of investor offers at an $800 billion valuation that the company has so far declined.
• CEO: Dario Amodei
• Other notable leadership: Daniela Amodei (President), Jared Kaplan (Chief Science Officer), Tom Brown, Jack Clark, Chris Olah, Sam McCandlish (all co-founders)
• Open weights: None. All Claude models are closed weights.
• Flagship models: Claude Opus 4.7 (April 2026), Claude 3.5 Sonnet, Claude 3 Haiku

Origins

Anthropic was founded in 2021 by a group of senior researchers and engineers who had recently left OpenAI. The lead founders were the Amodei siblings: Dario, who had been OpenAI's Vice President of Research, and Daniela, who had led safety and policy. They were joined by Jared Kaplan, Tom Brown (a lead author on the GPT-3 paper), Sam McCandlish, Chris Olah, Jack Clark, and several others. The departure was framed publicly as a need to refocus AI development around safety research with structural guarantees the OpenAI charter no longer provided.

Anthropic was incorporated as a Public Benefit Corporation, a Delaware structure that allows the company to balance fiduciary duty with a stated public benefit mission. The mission, codified in the founding documents, is "the responsible development and maintenance of advanced AI for the long-term benefit of humanity." The structure differs from OpenAI's capped-profit-under-nonprofit arrangement.

The first public Claude model launched in March 2023. The Claude product line, organized into Opus (largest), Sonnet (mid-tier), and Haiku (smallest and fastest) tiers, has expanded across multiple major version generations through to Claude Opus 4.7 in April 2026.

Mission and strategy

Anthropic's stated mission is to develop AI that is "steerable, interpretable, and safe." Constitutional AI (CAI), a methodology in which models are trained to follow a written constitution of principles via self-critique and revision, is the company's signature published research contribution and the basis for the alignment posture in shipped Claude models.

The commercial strategy emphasizes enterprise customers over consumer reach. Claude is sold through the Anthropic API directly and through cloud-partner channels: AWS Bedrock and Google Vertex AI. Self-serve Enterprise plans, expanded in April 2026 to allow direct purchase without a sales conversation, broadened the distribution further. Product extensions have followed the same enterprise-first pattern: Claude Code (a coding agent), Claude Cowork (a desktop app for Mac and Windows with role-based access controls), Claude Design (a visuals and prototyping product introduced in April 2026), and Managed Agents (a hosted agent platform).

The competitive premise is that frontier capability combined with safety differentiation and enterprise channel depth produces durable B2B revenue, even where capability margins against OpenAI are narrow. The April 2026 announcement that OpenAI was prioritizing enterprise sales is widely read as a direct response to Anthropic's enterprise traction.

Models and products

• Claude family. Claude Opus 4.7 (April 2026) is the current flagship at the largest tier, shipped after iterative point releases through 4.0, 4.5, and 4.6. Claude 3.5 Sonnet remains a mid-tier option used widely in production deployments. Claude 3 Haiku is the smallest and fastest variant. All closed weights.
• Claude Code. Coding agent introduced in 2025 as Anthropic's response to OpenAI's Codex and Cursor's Composer. Integrated with Claude Opus 4.7 and IDE tooling.
• Claude Cowork. Desktop app for macOS and Windows, generally available in April 2026. Adds enterprise admin controls, OpenTelemetry support, and role-based access for team and enterprise plans.
• Claude Design. Generative product for visuals, prototypes, slides, and one-pagers. Powered by Claude Opus 4.7. Released as research preview in April 2026 for Pro, Max, Team, and Enterprise tiers.
• Managed Agents. Hosted Claude Platform service for long-running agent workflows, with stable session interfaces, sandboxes, and durable state.
• Plugin marketplace. Launched alongside admin controls for team and enterprise plans in April 2026.

Anthropic also distributes Claude through AWS Bedrock and Google Vertex AI, and provides developer access via the Anthropic API.

Benchmarks and standing

As of April 2026, Claude Opus 4.7 sits at #2 across most major frontier benchmarks behind OpenAI's GPT-5.5, with one notable exception. On SWE-bench Verified, Anthropic's Claude 4.5 Opus leads at 74.0 against GPT-5.5 at 68.5, the only major coding benchmark where Anthropic holds the top position.

Specific April 2026 standings: Artificial Analysis Intelligence Index #2 at 57.28; LMArena General ELO #2 at 1298; LMArena Coding ELO #2 at 1398; LMArena Vision ELO #3 at 1287; GPQA Diamond #2 at 91.8%; HumanEval+ #2 at 93.8%; ARC-AGI Challenge #2 at 88.7; AIME 2025 #2 at 93.3%.

The pattern of consistent second-place finishes with a clear coding-benchmark lead reflects the product strategy: enterprise developers buying for software engineering work have a measurable case for Claude over GPT-5.5.

Leadership

As of April 2026, Anthropic's senior leadership includes:

• Dario Amodei, Chief Executive Officer and co-founder. Formerly OpenAI's Vice President of Research. Public face for the company on capability claims, safety positioning, and policy engagement.
• Daniela Amodei, President and co-founder. Formerly OpenAI's Vice President of Safety and Policy. Operational lead and outward face for partner relationships.
• Jared Kaplan, Chief Science Officer and co-founder. Theoretical physicist by training; co-author of the foundational scaling-laws paper that informs much of frontier-model training strategy.
• Tom Brown, co-founder. Lead author of the GPT-3 paper while at OpenAI; technical leadership on training and infrastructure.
• Jack Clark, co-founder. Policy lead and external-facing voice on AI governance.
• Chris Olah, co-founder. Interpretability research, including the development of mechanistic interpretability work that has shaped the broader field.
• Sam McCandlish, co-founder. Research leadership.

Notable additions include John Schulman, an OpenAI co-founder and reinforcement learning specialist who joined Anthropic in 2024. The leadership team has been comparatively stable; departures from the founding cohort have been rare.

Funding and backers

Anthropic has raised approximately $73 billion in cumulative funding across multiple rounds. The funding history includes a $124 million Series A in 2021, a $580 million Series B in 2022 (with Sam Bankman-Fried's Alameda Research participation later disputed in FTX bankruptcy proceedings), Google strategic investments totaling more than $2 billion through 2024, and Amazon commitments totaling $8 billion through 2024 with AWS as the primary compute partner.

Recent rounds have driven valuation rapidly higher: a $61.5 billion mark in March 2025, then a $13 billion Series F at $183 billion in September 2025, followed by a $30 billion Series G in February 2026 at $380 billion post-money, the second-largest private financing on record. The Series G was led by GIC and Coatue, with co-leads including D.E. Shaw Ventures, Dragoneer, Founders Fund, ICONIQ, and MGX.

In April 2026, Bloomberg and TechCrunch reported that Anthropic was receiving investor offers at an $800 billion valuation that the company had so far declined to act on. The reported reluctance has been characterized as a signal of confidence in the underlying revenue trajectory rather than a refusal to raise.

Industry position

Anthropic occupies the second position behind OpenAI across most frontier-capability metrics, with a leading position in coding benchmarks and a differentiated commercial channel. The capability gap on most benchmarks is small, often within five points on the Artificial Analysis Intelligence Index. Anthropic's distinguishing characteristic is the combination of safety-oriented research output and enterprise-first commercial focus.

The Series G at $380 billion in February 2026 places Anthropic among the most valuable private companies in the world, and the resistance to $800 billion offers in April 2026 suggests the company believes near-term growth will support an even higher mark.

Strategic risks identified in industry coverage include the narrowness of the capability gap to OpenAI (which constrains pricing power if the gap widens), the dependency on AWS and Google as both compute partners and competitors, and the durability of the enterprise wins as OpenAI publicly pivots to enterprise as its 2026 priority. Investor enthusiasm at the $800 billion mark indicates that current revenue is interpreted as supporting the valuation, but the path to public-company-style margins remains an open question across the Frontier category.

Competitive landscape

Anthropic competes with several Frontier and Insurgent labs:

• OpenAI. The primary head-to-head competitor on capability and the principal target of Anthropic's enterprise sales motion. OpenAI's GPT-5.5 leads most benchmarks but trails on SWE-bench Verified. OpenAI's announced 2026 enterprise priority is a direct response to Anthropic's traction.
• Google DeepMind. Gemini 3.1 Pro is benchmark-competitive and embedded into Google Search and Workspace, which is a different distribution model than Anthropic's API-and-cloud channel.
• DeepSeek. Chinese open-weights frontier capability at significantly lower per-token cost. The cost gap creates substitution risk for cost-sensitive enterprise customers.
• Mistral AI. European frontier lab with a different distribution profile (open weights, EU regulatory tailwind) but smaller scale.
• Cohere. Enterprise-focused frontier lab with a narrower product line and a more retrieval-and-RAG-oriented positioning.
• Insurgent labs. Several recent OpenAI departures founded competing labs (Thinking Machines Lab, Safe Superintelligence). None has shipped a competing public product as of April 2026 but each is watched as a potential future capability entrant.

Outlook

Several open questions affect Anthropic's trajectory in 2026 and 2027:

• The outcome of the $800 billion valuation offers: whether the company accepts a new round, holds out for a higher mark, or moves toward an IPO instead.
• The durability of Anthropic's enterprise lead as OpenAI publicly pursues the same customers and as Microsoft pushes Copilot adoption.
• The release timing and capability profile of Claude 5 or the next major-version step beyond the 4.x line.
• The status of Anthropic's compute relationships with AWS and Google, particularly any expansion of the Amazon commitment beyond the existing $8 billion.
• The trajectory of Constitutional AI methodology and whether competitors adopt similar safety-training techniques.
• Possible expansion or restructuring of Anthropic's PBC obligations as the company scales toward public-market candidacy.

Sources

• Anthropic: Introducing Claude Opus 4.7. Latest flagship release announcement.
• Anthropic: $30B Series G at $380B post-money valuation. February 2026 funding announcement.
• TechCrunch: Anthropic shrugs off VC offers valuing it at $800B+. April 2026 valuation context.
• Bloomberg: Anthropic Attracts Investor Offers at an $800 Billion Valuation. Investor demand reporting.
• TechCrunch: Anthropic launches Claude Design. April 2026 product launch.
• Artificial Analysis Intelligence Index. Current frontier benchmark standings.
• SWE-bench Verified leaderboard. Coding benchmark where Claude leads.