· May 22, 2025
Anthropic introduces Claude Opus 4 and Sonnet 4 for coding and longer tasks
Markdown · JSON · Original source
announced Opus 4 and Claude Sonnet 4 on May 22, 2025, describing them as models for coding, reasoning and . Anthropic says Opus 4 can sustain work on complex, long-running tasks, while Sonnet 4 improves on Sonnet 3.7 and offers a balance of performance and efficiency. Both models can use tools during extended thinking and in parallel; with local file access, they can save key facts to maintain continuity. Anthropic also announced Claude Code’s general availability, including background tasks and integrations with code editors, and new capabilities for developers building AI agents. Anthropic presents these releases as a step toward a “virtual collaborator” that can maintain context and focus on longer projects. Its case rests partly on results and customer reports, including Rakuten’s report of a refactor that ran independently for seven hours. Those examples describe tests and reported uses, not how reliably the models will perform across everyday work. Anthropic says the models underwent testing and evaluation intended to reduce risk and improve safety; the announcement does not establish their effects on users’ work beyond the examples it gives.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: May 22, 2025 (2025-05-22T00:00:00.000Z).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-14T00:33:25.890Z. First observed: 2026-10-22T07:14:19.956Z. Retrieved: 2026-10-24T02:49:35.518Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 24, 2026; originally published May 22, 2025. ’s announcement includes its own model claims, results and customer reports; the evidence does not independently establish how the models perform in everyday work. The reported examples do not establish performance across other tasks or users.
Open question
How reliably do Opus 4 and Sonnet 4 sustain useful work on long-running tasks outside the reported tests and customer examples?
Source-stated
Anthropic announced Claude Opus 4 and Claude Sonnet 4 as its next generation of models.
Today, we’re introducing the next generation of Claude models: Claude Opus 4 and Claude Sonnet 4
Source-stated
Anthropic says Opus 4 sustains performance on complex, long-running tasks and agent workflows.
Claude Opus 4 is the world’s best coding model, with sustained performance on complex, long-running tasks and agent workflows.
Source-stated
Anthropic says both models can use tools during extended thinking and use tools in parallel.
Both models can use tools—like web search—during extended thinking, allowing Claude to alternate between reasoning and tool use to improve responses.
Source-stated
Anthropic describes the releases as a step toward a virtual collaborator that can maintain context and focus on longer projects.
These models are a large step toward the virtual collaborator—maintaining full context, sustaining focus on longer projects, and driving transformational impact.
Source-stated
Anthropic reports that Rakuten tested Opus 4 on an open-source refactor that ran independently for seven hours.
Rakuten validated its capabilities with a demanding open-source refactor running independently for 7 hours with sustained performance.
Source-stated
Anthropic says it conducted testing and evaluation intended to minimize risk and maximize safety.
They come with extensive testing and evaluation to minimize risk and maximize safety
Source and contributor context
Anthropic · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
AI agent
An AI system that chooses and carries out actions toward a goal, often using software tools.
API
An application programming interface: a defined way for software to request data or actions from another system.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Editorial provenance
Version 1. Publication ID: e8768e33-b32e-2ee1-e193-b2fbff693c31.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-14T00:28:54.420Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: d9008fb2a2835958e35c87bec4ff1ef8c0c688f3b62bf904731f13eb5da82f99. Source text hash: 6abffff5a2a104a024eacb5320e90839620f91c60a395c1757e454ff5772d64f.
Revision and source-check history
Version 1 · 2026-11-14T00:33:25.890Z
Anthropic introduces Claude Opus 4 and Sonnet 4 for coding and longer tasks
announced Opus 4 and Claude Sonnet 4 on May 22, 2025, describing them as models for coding, reasoning and . Anthropic says Opus 4 can sustain work on complex, long-running tasks, while Sonnet 4 improves on Sonnet 3.7 and offers a balance of performance and efficiency. Both models can use tools during extended thinking and in parallel; with local file access, they can save key facts to maintain continuity. Anthropic also announced Claude Code’s general availability, including background tasks and integrations with code editors, and new capabilities for developers building AI agents. Anthropic presents these releases as a step toward a “virtual collaborator” that can maintain context and focus on longer projects. Its case rests partly on results and customer reports, including Rakuten’s report of a refactor that ran independently for seven hours. Those examples describe tests and reported uses, not how reliably the models will perform across everyday work. Anthropic says the models underwent testing and evaluation intended to reduce risk and improve safety; the announcement does not establish their effects on users’ work beyond the examples it gives.
Original source
· Article
Source details
Original source date: May 22, 2025 (2025-05-22T00:00:00.000Z).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-14T00:33:25.890Z. First observed: 2026-10-22T07:14:19.956Z. Retrieved: 2026-10-24T02:49:35.518Z.
Limits
Text as retrieved on October 24, 2026; originally published May 22, 2025. ’s announcement includes its own model claims, results and customer reports; the evidence does not independently establish how the models perform in everyday work. The reported examples do not establish performance across other tasks or users.
Open question
How reliably do Opus 4 and Sonnet 4 sustain useful work on long-running tasks outside the reported tests and customer examples?
Source-stated
Anthropic announced Claude Opus 4 and Claude Sonnet 4 as its next generation of models.
Today, we’re introducing the next generation of Claude models: Claude Opus 4 and Claude Sonnet 4
Source-stated
Anthropic says Opus 4 sustains performance on complex, long-running tasks and agent workflows.
Claude Opus 4 is the world’s best coding model, with sustained performance on complex, long-running tasks and agent workflows.
Source-stated
Anthropic says both models can use tools during extended thinking and use tools in parallel.
Both models can use tools—like web search—during extended thinking, allowing Claude to alternate between reasoning and tool use to improve responses.
Source-stated
Anthropic describes the releases as a step toward a virtual collaborator that can maintain context and focus on longer projects.
These models are a large step toward the virtual collaborator—maintaining full context, sustaining focus on longer projects, and driving transformational impact.
Source-stated
Anthropic reports that Rakuten tested Opus 4 on an open-source refactor that ran independently for seven hours.
Rakuten validated its capabilities with a demanding open-source refactor running independently for 7 hours with sustained performance.
Source-stated
Anthropic says it conducted testing and evaluation intended to minimize risk and maximize safety.
They come with extensive testing and evaluation to minimize risk and maximize safety
Source and contributor context
Anthropic · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
AI agent
An AI system that chooses and carries out actions toward a goal, often using software tools.
API
An application programming interface: a defined way for software to request data or actions from another system.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Editorial provenance
Version 1. Publication ID: e8768e33-b32e-2ee1-e193-b2fbff693c31.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-14T00:28:54.420Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: d9008fb2a2835958e35c87bec4ff1ef8c0c688f3b62bf904731f13eb5da82f99. Source text hash: 6abffff5a2a104a024eacb5320e90839620f91c60a395c1757e454ff5772d64f.
passed · 2026-11-14T00:28:54.420Z · anthropic/claude-opus-5.5
Publication: e8768e33-b32e-2ee1-e193-b2fbff693c31
attribution — pass: The reading attributes the model claims, benchmark framing, customer reports and safety statements to Anthropic, the publisher of this single announcement. Customer examples such as Rakuten's are presented as reported by Anthropic, not as established fact.
Today, we’re introducing the next generation of Claude models: Claude Opus 4 and Claude Sonnet 4
summary — pass: The reading leads with the actor and the development, then explains the significance Anthropic claims: a step toward a virtual collaborator. It also notes the limits of tests and customer examples. AI agent, API and benchmark are defined beside the reading. 'Extended thinking' and 'refactor' have no tooltips; adding term-library entries is suggested. Context makes them inferable, and they do not block understanding of the central development.
These models are a large step toward the virtual collaborator—maintaining full context, sustaining focus on longer projects, and driving transformational impact.
date — pass: May 22, 2025 matches the UTC calendar day of originalPublished, an instant. The page was modified on 2026-10-14. The cited passages are core announcement text rather than plausible later additions. There is no hindsight framing, and no forecast is presented as an outcome. The evidence is a publisher page, so no archive disclosure is required.
Today, we’re introducing the next generation of Claude models
support — pass: Each claim matches its quoted source passage, including the claims on tool use, long-running tasks, Rakuten's seven-hour refactor, Claude Code availability and safety testing. The reading does not turn benchmark or customer reports into demonstrated real-world reliability. The question follows from that gap without leading. Although the page was edited after publication, the cited passages are plainly part of the core announcement.
Rakuten validated its capabilities with a demanding open-source refactor running independently for 7 hours with sustained performance.
media — pass: No media was selected, and no verified media candidates were supplied, so no media identity or relevance issue arises.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceAn application programming interface: a defined way for software to request data or actions from another system.
Definition referenceA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceAn application programming interface: a defined way for software to request data or actions from another system.
Definition referenceA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore Claude