The sleeping giant has finally awoken, as Anthropic has released the highly anticipated Claude 4 series. This new generation of models, Claude 4 Opus and Claude 4 Sonnet, are setting new standards in coding, reasoning, and agentic workflows.

Claude 4 Opus is the flagship model, topping the Sway Bench at 72.5% and the Terminal Bench at 43.2%. It excels at long-running tasks, showcasing sustained focus for hours. Opus boasts deep multifile code understanding, editing, and debugging capabilities, and even maintains memory across tasks, allowing it to create a navigation guide while playing Pokémon.

Claude 4 Sonnet, on the other hand, strikes the perfect balance of performance and speed. It features a hybrid mode that allows you to switch between instant replies and extended thinking for deep reasoning. Sonnet also introduces tool use, parallel tool execution, improved memory from file storage, and a new Cloud Code tool with native VS Code and JetBrains extensions.

Both models outperform the competition, including OpenAI's Codex 1, the O3, and GPT-4.1, in various coding benchmarks. The Opus 4 reaches an impressive 79.4% accuracy, while the Sonnet 4 tops the chart at 80.2%.

These models are the most capable for complex software engineering workflows, making them ideal for developers, researchers, and power users who require high performance and full control through the new developer mode. They excel at end-to-end app generation, long-context workflows, and prompt engineering.

To get started with these models, you can access them through Anthropic's chatbot, console, API, or OpenRouter. Whether you're looking to build autonomous agents, tackle multi-step tasks, or create tools that require long-term coherence, the Claude 4 series is the solution you've been waiting for.

The Claude 4 Opus and Claude 4 Sonnet models from Anthropic have set new standards in coding, reasoning, and agentic workflows. The Claude 4 Opus, the flagship model, excels at complex reasoning, coding, and agentic tasks. It boasts impressive benchmark scores, topping the Sway Bench at 72.5% and the Terminal Bench at 43.2%.

The key features that set the Claude 4 Opus apart include reliable long-term reasoning, advanced memory through local file access, and thinking summaries with a developer mode. These capabilities allow the model to maintain context and continuity over long tasks, making it ideal for building autonomous agents, multi-step workflows, and tools that require long-term coherence.

The Claude 4 Sonnet, a smaller and faster sibling to the Opus model, also offers impressive performance. It features a hybrid thinking mode, allowing for both instant replies and deeper reasoning when needed. The Sonnet shares key improvements with the Opus, such as reduced shortcut behaviors, tool use, and reasoning tool switching. It provides solid performance at a lower latency and cost, making it a compelling option for developers, researchers, and power users.

Both the Claude 4 Opus and Claude 4 Sonnet have demonstrated their capabilities in various coding tasks, from building responsive web applications to generating creative Tetris games. The models' ability to generate high-quality code, maintain context, and execute complex workflows showcases Anthropic's advancements in the field of AI-powered coding and reasoning.

The Claude 4 Opus and Claude 4 Sonnet models introduced by Anthropic represent significant advancements in coding, reasoning, and agentic workflows.

The Claude 4 Opus is Anthropic's flagship model, excelling at complex reasoning, coding, and autonomous agent tasks. It supports extended memory through local file access, allowing it to maintain context and continuity over long workflows. This makes it ideal for building end-to-end applications, executing multi-step tasks, and prompt engineering. Key enhancements include reliable long-term reasoning, advanced memory management, and developer-friendly thinking summaries.

The Claude 4 Sonnet is a smaller, faster sibling to the Opus model, offering a balance of performance and speed. It features a hybrid thinking mode, allowing users to switch between instant replies and deeper reasoning as needed. The Sonnet also shares improvements like reduced shortcut behaviors, tool use, and reasoning tool switching, providing solid performance at a lower latency and cost.

Both models leverage Anthropic's latest advancements, including cloud code, parallel tool execution, and enhanced memory and API capabilities. These features position the Claude 4 models as the most capable options for complex software engineering workflows, outperforming competing models in various coding, math, and agentic benchmarks.

The Claude 4 Opus and Sonnet models are the latest releases from Anthropic, showcasing their dominance in software engineering tasks. The Opus model is the flagship, excelling at complex reasoning, coding, and agentic workflows. It supports extended memory through local file access, allowing it to maintain context and continuity over long tasks. This makes it ideal for building autonomous agents, multi-step tasks, and tools requiring long-term coherence. The Opus model is best suited for developers, researchers, and power users who need high performance and full control through its new developer mode.

The Claude 4 Sonnet, on the other hand, is a smaller and faster sibling to the Opus model. It features a hybrid thinking mode, allowing for instant replies or deeper reasoning when needed. The Sonnet shares key improvements with the Opus, such as reduced shortcut behaviors, tool use, and reasoning tool switching. It offers solid performance at a low latency and cost, making it a more accessible option.

Both models have demonstrated impressive capabilities in various coding tasks, such as building a responsive web page for personal finance tracking, creating a TV channel simulator with creative animations, and generating an SVG representation of a butterfly. The Opus model consistently outperformed the Sonnet in terms of coherence, execution, and attention to detail, showcasing its superior capabilities for complex software engineering workflows.

In terms of pricing, the Claude 4 Opus is significantly more expensive, with an input price of $15 per 1 million tokens and an output price of $75 per 1 million tokens. The Claude 4 Sonnet, on the other hand, uses the same pricing as the Claude 3.7, with an input price of $3 per 1 million tokens and an output price of $15 per 1 million tokens.

Overall, the Claude 4 Opus and Sonnet models represent a significant advancement in Anthropic's capabilities, positioning them as the go-to choices for developers, researchers, and power users who require high-performance, long-term reasoning, and coherence in their coding and agentic workflows.

To assess the capabilities of the Claude 4 Opus and Claude 4 Sonnet models, we put them through a series of coding challenges. Here's how they performed:

Personal Finance Tracking App

Both models were tasked with building a responsive web page using HTML, CSS, and JavaScript that allows users to track their monthly income and expenses. The Opus model excelled, generating a comprehensive application with features like adding, editing, and deleting transactions, as well as a night mode and export functionality. The Sonnet model also performed well, replicating the overall style, but with some missing components compared to the Opus version.

TV Channel Simulator

For this challenge, the models were asked to create a TV channel simulator with 10 different channels (0-9), each with unique animations and visuals using the p5.js library. The Opus model delivered a polished, static animation when switching between channels, while the Sonnet model generated a more creative set of channel designs, including a sports zone, weather channel, and cosmic TV.

Butterfly SVG

Both models were instructed to create an SVG representation of a butterfly with symmetrical wings and simple styling. The Opus model excelled, producing a beautiful and accurate butterfly design. The Sonnet model, while stylistically pleasing, did not quite capture the essence of a butterfly, resembling more of a worm-like creature.

Tetris Game

As a final test, the models were challenged to create a functional Tetris game with a score display and controls. The Opus model outshone the Sonnet, generating a fully animated Tetris game with a working scoreboard and smooth block mechanics.

Overall, the Claude 4 Opus model demonstrated superior capabilities in complex coding tasks, showcasing its reliable long-term reasoning, advanced memory, and thoughtful summaries. The Sonnet model also performed well, striking a balance between speed and performance. Both models proved to be impressive in their coding abilities, with the Opus model emerging as the clear choice for developers, researchers, and power users who require high-performance and precise control.

The Claude 4 Opus and Claude 4 Sonnet models from Anthropic have set new standards in coding, reasoning, and agentic workflows. The Opus model, in particular, excels at complex reasoning, coding, and agentic tasks, with its reliable long-term reasoning, advanced memory through local file access, and thinking summaries with a developer mode. It is ideal for building autonomous agents, multi-step tasks, and tools that require long-term coherence.

The Sonnet model, on the other hand, offers a balance of performance and speed, with a hybrid thinking mode that allows for instant replies or deeper reasoning when needed. It shares key improvements with the Opus model, such as reduced shortcut behaviors and tool use and reasoning tool switching.

Both models have demonstrated impressive capabilities in various coding tasks, from building a responsive web page for personal finance tracking to generating a Tetris game with animations and a scorecard. The Opus model, in particular, has shown its superiority in terms of the quality and complexity of the generated code.

While the pricing for the Opus model may be considered high, with an input price of $15 per 1 million tokens and an output price of $75 per 1 million tokens, the model's capabilities make it a valuable investment for developers, researchers, and power users who require high-performance and full control through the developer mode.

Overall, the release of the Claude 4 Opus and Claude 4 Sonnet models by Anthropic is a significant step forward in the field of coding and agentic workflows, and they are poised to become the go-to models for complex software engineering tasks.