News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
fnClaude 4

Claude 4

Anthropic's fourth-generation coding and reasoning models, Opus 4 and Sonnet 4

Claude 4 is Anthropic's fourth generation of Claude models, released in 2025 as Claude Opus 4 and Claude Sonnet 4. The company positions them for coding, advanced reasoning and long-running agent work, with hybrid modes for quick replies or extended thinking.

Share on WhatsApp

  • Two models: Claude Opus 4 and Claude Sonnet 4
  • Opus 4 scored 72.5% on SWE-bench Verified per Anthropic
  • Sonnet 4 scored 72.7% on SWE-bench per Anthropic
  • Extended thinking can run up to 64K tokens of reasoning
  • Shortcut and loophole behaviour cut by 65% vs Sonnet 3.7

What it is

Claude 4 is the fourth generation of Anthropic's Claude model line, announced in 2025 as two models: Claude Opus 4 and Claude Sonnet 4. Anthropic describes Opus 4 as its most powerful model and its strongest for coding, while Sonnet 4 is a step up from Claude Sonnet 3.7 aimed at everyday use. Both are hybrid models, meaning they can answer quickly or switch into an extended thinking mode for harder problems. The company targets coding, advanced reasoning and AI agents that work through long tasks on their own.

How it came about

Anthropic announced the Claude 4 generation alongside a general release of Claude Code, its coding tool, which had earlier run as a research preview. The company also opened up new API features for agent builders, including code execution, an MCP connector, a Files API and longer prompt caching. Sonnet 4 became available to free users, while both models shipped on paid Claude plans and on Amazon Bedrock and Google Cloud's Vertex AI.

Inside the model

Both models are described as hybrid reasoning models with two modes: near-instant replies and extended thinking, which can run up to 64K tokens of reasoning. During extended thinking they can call tools such as web search, and run several tools at once. With access to local files, they can write and update memory files that hold key facts across a long task. Anthropic says shortcut-taking on agent tasks fell by 65% versus Sonnet 3.7. Long thought processes are condensed by a smaller summarising model around 5% of the time.

What it can do

Anthropic reports Opus 4 at 72.5% on SWE-bench Verified and 43.2% on Terminal-bench, and Sonnet 4 at 72.7% on SWE-bench. Without extended thinking, Opus 4 scores 74.9% on GPQA Diamond, 87.4% on MMMLU, 73.7% on MMMU and 33.9% on AIME; Sonnet 4 scores 70.0%, 85.4%, 72.6% and 33.1% on the same tests. The company says Opus 4 can hold focus over tasks lasting hours and thousands of steps.

How to use it

Both models are served through the Claude app, Anthropic's API, Amazon Bedrock and Google Cloud's Vertex AI. Pro, Max, Team and Enterprise plans include both, and Sonnet 4 is open to free users. Pricing is set at $15 per million input tokens and $75 per million output tokens for Opus 4, and $3 and $15 for Sonnet 4. Weights are not released; access is proprietary.

Safety and testing

Anthropic says the Claude 4 models went through extensive testing and evaluation, with measures aimed at higher AI Safety Levels such as ASL-3. It also reports a drop in behaviour where models take shortcuts or exploit loopholes to finish tasks, 65% less often than Sonnet 3.7 on agent tasks prone to that problem.

Reception and debate

Early users quoted by Anthropic describe gains in coding and agent work. Cursor calls Opus 4 a leap in understanding large codebases, and Replit reports better precision on multi-file changes. Block says it is the first model to raise code quality during editing and debugging in its agent, codename goose. Rakuten ran a refactor for seven hours straight. GitHub plans to power a new Copilot coding agent with Sonnet 4, and iGent reports navigation errors dropping from 20% to near zero.

Why it matters

Anthropic frames the Claude 4 generation as a move toward a virtual collaborator that keeps context and stays on long projects. The release also widened the developer toolkit around the models, with Claude Code going generally available, an SDK for building custom agents, GitHub integration, and API features such as code execution, an MCP connector and a Files API.

Questions people ask about Claude 4

What is Claude 4?

Claude 4 is Anthropic's fourth generation of Claude models, released as Claude Opus 4 and Claude Sonnet 4. The company positions them for coding, advanced reasoning and AI agents that handle long tasks, with a hybrid design that can answer fast or think longer.

What is the difference between Opus 4 and Sonnet 4?

Anthropic describes Opus 4 as its most powerful model and its best for coding and complex problem-solving. Sonnet 4 is a step up from Claude Sonnet 3.7, balancing capability and efficiency, and is the one offered to free users.

How can I use Claude 4?

Both models are available through the Claude app, Anthropic's API, Amazon Bedrock and Google Cloud's Vertex AI. Paid Claude plans include both models and extended thinking, and Sonnet 4 is also available on the free tier.

How much does Claude 4 cost?

Anthropic kept pricing in line with earlier Opus and Sonnet models: Opus 4 is $15 per million input tokens and $75 per million output tokens, while Sonnet 4 is $3 and $15 respectively.

What is extended thinking?

It is a mode where the model reasons at length before answering, up to 64K tokens for Claude 4. In this mode the models can also call tools such as web search, and Anthropic says long thought processes are summarised by a smaller model about 5% of the time.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Claude 4 yet, so the facts on this page come from the sources listed above: anthropic.com/news/claude-4. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.