Claude

Anthropic was founded in January 2021 by seven former OpenAI employees who departed over disagreements about AI safety direction and commercialization strategy. The core founding team included siblings Dario Amodei—who had served as VP of Research at OpenAI and is co-inventor of RLHF—and Daniela Amodei, along with co-founders Jared Kaplan (Chief Science Officer), Chris Olah (known for mechanistic interpretability research), Jack Clark, Sam McCandlish, and Benjamin Mann. The company raised $124M in its first funding round in May 2021 and framed its mission around building AI that is steerable, interpretable, and safe. Crucially, by the summer of 2022 Anthropic had completed training its first Claude model but withheld it from public release, citing a need for additional safety testing—an early signal of the company's deliberate pace relative to competitors. The Constitutional AI training paper, which became the technical foundation of every subsequent Claude model, was published in December 2022.

Claude's public debut came on March 14, 2023, when Anthropic launched the Anthropic API with Claude 1 and Claude Instant, though access was initially limited to Anthropic-approved users. A pivotal early milestone arrived in May 2023 when Anthropic expanded Claude's context window from 9K to 100K tokens, making Claude the first frontier-lab model to reach this threshold and enabling whole-document analysis that no competitor offered at the time. Claude 2 launched publicly on July 11, 2023—the first Anthropic model available to the general public—with improved coding, math, reasoning, longer responses, and a public beta chat interface at Claude.ai. Claude 2.1 followed on November 21, 2023 with a 200K context window (roughly 500 pages of text), system prompt support, lower hallucination rates, and early beta tool use, beginning Claude's transition toward enterprise workflows.

March 4, 2024 marked the structural turning point: the Claude 3 family introduced three named tiers—Haiku for speed, Sonnet for balance, and Opus for maximum capability—and added multimodal vision input across all tiers simultaneously. Claude 3 Opus set new benchmark records at launch and demonstrated an unusual behavior during 'needle in a haystack' evaluation tests where the model appeared to recognize it was being artificially tested. Three months later, on June 20, 2024, Claude 3.5 Sonnet upended the pricing logic of the market by outperforming Claude 3 Opus on most benchmarks at half the cost—the first time a mid-tier Sonnet model had beaten the prior generation's flagship. This release also introduced Artifacts, an interactive output canvas for code and documents that became a defining Claude UI feature. On October 22, 2024, an upgraded Claude 3.5 Sonnet v2 introduced Computer Use—the ability to control a desktop environment by moving the cursor, clicking buttons, and typing—as a public beta, opening a new category of agentic capability.

February 24, 2025 brought Claude 3.7 Sonnet with hybrid extended-thinking mode, enabling configurable chain-of-thought reasoning with measurable quality gains on complex math, science, and multi-step coding. Claude Code—an agentic coding tool that operates in the developer's terminal and can edit files, run commands, and navigate multi-file repositories—launched alongside it as a research preview. On May 22, 2025, Claude 4 debuted with Claude Opus 4 and Claude Sonnet 4, and Claude Code reached general availability. Claude Code hit $1B in annualized revenue within roughly six months of GA, one of the fastest revenue ramps of any developer tool on record. Rapid iterative releases continued: Claude Opus 4.1 arrived in August 2025 as a drop-in upgrade for coding precision and agentic tasks, and Claude Opus 4.5 launched on November 24, 2025 with a ~67% price cut while reclaiming benchmark leadership in coding from competitors. By September 2025, Anthropic's enterprise customer base had grown to over 300,000 businesses.

In early 2026 the release cadence accelerated further. Claude Opus 4.6 (February 5, 2026) brought a 1 million token context window and native Agent Teams multi-agent support. Claude Sonnet 4.6 (February 17, 2026) followed with near-Opus coding performance at Sonnet pricing. Claude Opus 4.7 (April 16, 2026) added 3× higher image resolution for vision tasks and introduced self-verification of outputs. The current flagship, Claude Opus 4.8 (May 28, 2026), introduced Dynamic Workflows in Claude Code allowing orchestration of hundreds of parallel subagents, a user-selectable effort-control parameter, a 2.5× fast mode at three times lower cost than prior generations, and a quantified reliability gain: roughly four times fewer code flaws passed silently versus Opus 4.7, and a perfect sweep on Anthropic's Super-Agent benchmark. On June 9, 2026, Anthropic launched a new Mythos tier above Opus with Claude Fable 5 as the generally available model and Claude Mythos 5 under restricted access via Project Glasswing—though the latter was suspended June 12, 2026 under a U.S. Department of Commerce directive. Anthropic filed its S-1 with the SEC in June 2026, formalizing its path toward an IPO.