Claude (AI model)

After several months of closed testing with partners such as Notion, Quora and DuckDuckGo, Anthropic made Claude more broadly available on March 14, 2023, through a chat interface and an API in its developer console. Anthropic described Claude as based on its research into training helpful, honest and harmless AI systems and suited to tasks such as summarization, search, writing, Q&A and coding; it launched in two versions: the high-performance Claude and the lighter, less expensive and much faster Claude Instant.⁠[1][1][1][1][1]

When the Amazon Bedrock service launched in April 2023, Claude 1.3 and Claude Instant were among the first foundation models available to AWS customers.⁠[2]

On May 9, 2023, Anthropic published Claude’s “constitution,” explaining its Constitutional AI approach of giving a language model explicit values set out in a constitution rather than values determined only implicitly through large-scale human feedback. On January 21, 2026, Anthropic published a new version of the constitution.⁠[3][3][3]

On July 11, 2023, Anthropic released Claude 2, available through the API and a new public beta website, claude.ai, which users in the United States and United Kingdom could try that day. Claude 2 raised the input limit per prompt to 100,000 tokens; Anthropic said it scored 76.5% on the multiple-choice section of the US bar exam, up from 73.0% for Claude 1.3. In August of that year, Claude 2 came to Amazon Bedrock.⁠[4][4][4][4][2]

On August 9, 2023, Anthropic released Claude Instant 1.2 through its API, saying it incorporated the strengths of Claude 2 in real-world use cases and showed significant gains in math, coding, reasoning and safety.⁠[5][5]

On September 7, 2023, Anthropic introduced Claude Pro, a paid plan for claude.ai initially available in the United States and United Kingdom for $20 or £18 a month, with five times the usage of the free tier.⁠[6][6]

Claude 2.1, released on November 21, 2023, expanded the context window to 200,000 tokens (about 150,000 words), reduced hallucination rates, and added system prompts and a beta tool-use feature.⁠[7][7][7]

On March 4, 2024, Anthropic released the Claude 3 family, consisting of three models in ascending order of capability: Claude 3 Haiku, Claude 3 Sonnet and Claude 3 Opus; Haiku, Sonnet and Opus have since been Claude’s standing tier names. Opus and Sonnet became available that day on claude.ai and the API, which was made generally available in 159 countries; Opus was priced at $15 per million input tokens and $75 per million output tokens, and all three models launched with a 200,000-token context window. Claude 3 Haiku followed on March 13.⁠[8][8][8][8][8][9]

On May 1, 2024, Anthropic introduced the Team plan for teams at $30 per user per month and released a free Claude iOS app.⁠[10][10][10]

On June 21, 2024, Anthropic released Claude 3.5 Sonnet, the first model in the Claude 3.5 family, available to free users and priced on the API at $3 per million input tokens and $15 per million output tokens. The same day, Claude.ai introduced Artifacts, a feature that displays generated code, documents and other content separately alongside the conversation.⁠[11][11][11][11]

On October 22, 2024, Anthropic released an upgraded Claude 3.5 Sonnet and a new model, Claude 3.5 Haiku, and introduced “computer use” in public beta, letting developers direct Claude through the API to look at a screen, move a cursor, click buttons and type text. Anthropic acknowledged that the feature was still experimental and at times cumbersome and error-prone. On December 3, the price of 3.5 Haiku was revised to $0.80 per million input tokens and $4 per million output tokens.⁠[12][12][12][12][12]

On February 24, 2025, Anthropic released Claude 3.7 Sonnet, which it called the first hybrid reasoning model on the market: it can answer almost instantly or carry out extended, step-by-step thinking visible to the user, at the unchanged price of $3/$15. It also introduced Claude Code, a command-line tool for agentic coding, as a limited research preview.⁠[13][13][13][13]

On May 22, 2025, Anthropic released Claude Opus 4 and Claude Sonnet 4, priced at $15/$75 and $3/$15 per million tokens respectively. Anthropic said Opus 4 scored 72.5% on SWE-bench; Claude Code became generally available at the same time.⁠[14][14][14][14]

Claude Opus 4.1, released on August 5, 2025, was an upgrade to Opus 4 that scored 74.5% on SWE-bench Verified, at the same price as Opus 4.⁠[15][15][15]

On September 29, 2025, Anthropic released Claude Sonnet 4.5, added checkpoints to Claude Code, brought code execution and file creation into the Claude apps, and opened the building blocks used to make Claude Code to developers as the Claude Agent SDK.⁠[16][16][16][16]

Claude Haiku 4.5, released on October 15, 2025, was a new small model available to all users, priced at $1 per million input tokens and $5 per million output tokens.⁠[17][17]

Claude Opus 4.5 (API model claude-opus-4-5-20251101), released on November 24, 2025, cut Opus-tier pricing to $5 per million input tokens and $25 per million output tokens.⁠[18][18]

Claude Opus 4.6, released on February 5, 2026, was the first Opus-class model with a 1-million-token context window (in beta), keeping the $5/$25 price; Anthropic said it could multitask autonomously within Cowork. Claude Sonnet 4.6, released on February 17, likewise offered a 1-million-token context window in beta and became the default model for the Free and Pro plans in claude.ai and Claude Cowork.⁠[19][19][19][20][20]

On April 7, 2026, Anthropic announced Project Glasswing, bringing together Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChaseJPMorgan Chase & Co.JPMorgan Chase & Co. is a financial holding company headquartered in New York that operates investment banking, consumer and small business financial services, commercial banking, financial transaction processing and asset management under its J.P. Morgan and Chase brands; its roots trace back to the Manhattan Company of 1799.Open the full entry, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks to secure critical software using its unreleased frontier model Claude Mythos Preview. Anthropic said Mythos Preview could surpass all but the most skilled humans at finding and exploiting software vulnerabilities and had already found thousands of high-severity vulnerabilities; the company committed up to $100 million in Mythos Preview usage credits for these efforts and $4 million in direct donations to open-source security organizations.⁠[21][21][21][21][21][21]

Claude Opus 4.7, released on April 16, 2026, kept the $5/$25 price. Anthropic said that, as planned under Project Glasswing, it would test new cybersecurity safeguards on less capable models first, and Opus 4.7 was the first such model, with cyber capabilities less advanced than Mythos Preview’s.⁠[22][22][22]

Claude Opus 4.8, released on May 28, 2026, kept the same price; at the same time, claude.ai users gained control over how much effort Claude puts into a task, Claude Code added “dynamic workflows,” and fast mode for Opus 4.8 (at 2.5 times the speed) became three times cheaper than for previous models.⁠[23][23][23][23][23]

On June 9, 2026, Anthropic released Claude Fable 5 and Claude Mythos 5, which share the same underlying model. Fable 5 was described as a Mythos-class model made safe for general use, with strong safeguards under which requests on some topics are answered by Claude Opus 4.8 instead, triggering on average in less than 5% of sessions; Mythos 5, with fewer safeguards, was released only to a small number of trusted Project Glasswing partners for defensive cybersecurity work. Both were priced at $10 per million input tokens and $50 per million output tokens, less than half the price of Mythos Preview.⁠[24][25][25][25][24][25]

At 5:21 p.m. Eastern Time on June 12, 2026, Anthropic received an export control directive from the US government, citing national security authorities, to suspend all access to Fable 5 and Mythos 5 by any foreign national, including foreign national Anthropic employees. Anthropic said that because it could not verify nationality in real time, it had to disable the two models for all customers, while other models were unaffected; the directive letter did not provide specific details of the national security concern.⁠[26][26][24][26][26]

Anthropic said it was complying with the government’s legal directive but disagreed that the finding of a narrow potential “jailbreak” should be cause for recalling a deployed commercial model; it said the government should have the ability to block unsafe deployments, but through a statutory process that is transparent, fair, clear and grounded in technical facts, and that this action did not adhere to those principles.⁠[26][26][26]

Claude Sonnet 5, released on June 30, 2026, became available across all plans and the default model for the Free and Pro plans, priced at $2 per million input tokens and $10 per million output tokens; Anthropic said its performance was close to that of Opus 4.8.⁠[27][27][27][27]

Anthropic later explained that the directive followed the government learning of a report by Amazon researchers whose prompting method bypassed Fable 5’s safeguards and got it to identify several software vulnerabilities, in one case also producing exploit code. After reviewing the report with the government, Amazon and others, Anthropic said less capable models such as Claude Opus 4.8 and GPT-5.5 could identify the same vulnerabilities, and it trained an improved safety classifier with the government, with blocked requests routed to Opus 4.8 instead.⁠[24][24][24][24][24]

On June 30, 2026, the export controls on Fable 5 and Mythos 5 were lifted, and Fable 5 was restored to users globally on the Claude Platform, Claude.ai, Claude Code and Claude Cowork from July 1. Access to Mythos 5 was first restored for a set of US organizations following the US government’s approval on June 26.⁠[24][24][24]

Claude Opus 5, released on July 24, 2026, was priced at $5 per million input tokens and $25 per million output tokens, the same as Opus 4.8; Anthropic said it came close to the frontier intelligence of Fable 5 at half the price, and it became the default model on Claude Max and the strongest model on Claude Pro.⁠[28][28][28][28]

On September 1, 2026, Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1, the same model with different levels of safeguards: Fable 5.1 is generally available, while Mythos 5.1 is offered only through trusted access programs, with safeguards designed for cybersecurity and life sciences work. Fable 5.1 cut the price of cache reads by 75% and otherwise kept Fable 5’s pricing of $10 per million input tokens and $50 per million output tokens; it can be used to discover software vulnerabilities but not to develop exploits for them.⁠[29][30][31][31][31][31]

On September 22, 2026, Anthropic introduced Claude Opus 5.5, the first model in the Claude 5.5 family, saying it performs at the level of Fable 5.1 on most work and costs 40% less to run than Opus 5, with prices of $4 per million input tokens and $20 per million output tokens, 20% less than Opus 5. It was Anthropic’s first release since it called for pacing the frontier and was tested before release by external evaluators including Frontier Design and METR; because its biology and cybersecurity capabilities are comparable to Mythos 5.1’s, Anthropic deployed it with safeguards similar to those on Fable 5.1.⁠[32][32][32][32]

Claude Sonnet 5.5, released on September 28, 2026, is the second model in the 5.5 family; it runs more than 30% faster than Sonnet 5 and is priced the same as Sonnet 5, at $2 per million input tokens and $10 per million output tokens.⁠[33][33]

On October 2, 2026, ArtCraftArtCraftArtCraft is an open-source AI image and video creation tool founded by Brandon Thomas, with its team based in Atlanta. It grew out of the AI creative company Storyteller and uses 2D and 3D canvases, character posing and scene blocking to control generated results. In October 2026, ArtCraft launched Crafting Apps, a series of open-source applications written in Rust and modeled on major Adobe and Microsoft software; Thomas said they were developed with the help of Claude Opus 5.5, and the project quickly drew attention as well as controversy over its completeness, copyright and security.Open the full entry founder Brandon ThomasBrandon ThomasBrandon Thomas is a software engineer, entrepreneur and independent filmmaker based in Atlanta who formerly worked as an engineer at Square and founded the speech synthesis website FakeYou, the AI creative company Storyteller.ai and the open-source AI image and video tool ArtCraft. In October 2026, under the ArtCraft name, he released seven open-source Rust applications modeled on Adobe’s main programs, saying they were written with the help of Claude Opus 5.5, drawing attention and controversy.Open the full entry said on Hacker News that he was using Opus 5.5 to write pure-Rust open-source implementations of Adobe’s entire software suite.⁠[34]

On October 6, 2026, Anthropic expanded its Cyber Verification Program, under which qualifying security professionals can apply to use advanced cyber capabilities on Claude Opus 5.5, Claude Sonnet 5.5, Claude Mythos 5.1 and future models. Claude Haiku 5.5, released on October 7, costs on average about 75% less to run than Haiku 4.5, priced at $0.10 per million input tokens and $0.50 per million output tokens for prompts of up to 100,000 tokens.⁠[21][21][35][35][35]

On October 8, 2026, Anthropic launched Claude Dashboards and Claude Motion in beta and announced that Docs, Slides and Design were out of beta and available on every plan. Claude Motion, for the Team and Enterprise plans, has Claude write code that turns text, charts, shapes and images into editable animations without using a video generation model; the result can be downloaded as an MP4 or edited further in tools such as AdobeAdobeAdobe (Adobe Inc.) is a software company headquartered in San Jose, California, founded in 1982 by John Warnock and Charles Geschke. Known for PostScript, PDF and creative software such as Acrobat, Photoshop, Illustrator, Premiere Pro and After Effects, it moved its main creative products to Creative Cloud subscriptions from 2013 and expanded into digital marketing and generative AI; it is listed on Nasdaq under the ticker ADBE.Open the full entry, Higgsfield and Runway. The same day, HiggsfieldHiggsfieldHiggsfield is an AI visual-content company founded in 2023 and headquartered in San Francisco. Beginning with the Diffuse mobile app, it developed a browser-based production platform for creators, marketing teams and enterprises, combining proprietary and third-party models for images, video, storyboards, character consistency and multi-step creative work.Open the full entry launched Katana, a video editing skill powered by Claude Motion and available in Claude through the Higgsfield MCP.⁠[36][36][36][36][36][37][37]

As of October 2026, the current models listed in the Claude Platform documentation were Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5, all supporting text and image input, a 1-million-token context window and up to 128,000 output tokens; the documentation recommends starting with Opus 5.5 for most work and using Fable 5.1 for more demanding long-horizon reasoning and agentic work. The Claude Cowork product page also showed that Claude Cowork had been folded into Claude itself and was rolling out to Pro and Max users.⁠[38][38][38][38][39]

Anthropic regularly retires older models and commits to giving at least 60 days’ notice for publicly released models. According to its documentation, Claude 1 and Claude Instant were retired on November 6, 2024; Claude 2, Claude 2.1 and Claude Sonnet 3 on July 21, 2025; Claude Opus 3 on January 5, 2026; Claude Sonnet 3.7 and Claude Haiku 3.5 on February 19, 2026; Claude Opus 4 and Claude Sonnet 4 on June 15, 2026; and Claude Opus 4.1 on August 5, 2026; Claude Sonnet 4.5 is scheduled to retire on November 30, 2026, with Claude Sonnet 5.5 as the recommended replacement. Anthropic says it has committed to long-term preservation of model weights.⁠[40][40][40][40][40][40][40][40][40][40][40][40][40]

Claude: Official website: https://claude.ai/ (opens in a new window).⁠[41]

Anthropic: Official website: https://www.anthropic.com/ (opens in a new window).⁠[42]

X: Public account: https://x.com/claudeai (opens in a new window)⁠[41]

X: Public account: https://x.com/AnthropicAI (opens in a new window)⁠[42]