Anthropic

Anthropic released Claude Sonnet 4.6, a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. The model features a 1M token context window in beta and approaches Opus-level intelligence at a lower price point. Users preferred Sonnet 4.6 over Sonnet 4.5 roughly 70% of the time in Claude Code testing, and even preferred it to Opus 4.5 59% of the time due to less overengineering and better instruction following.

• Approaches Opus-level intelligence
• 1M token context window (beta)
• Significantly improved computer use skills
• Better instruction following and consistency
• Pricing remains $3/$15 per million tokens

Anthropic released Claude Opus 4.6, their most intelligent model with industry-leading performance on agentic coding, computer use, tool use, search, and finance. The model achieved state-of-the-art scores on Terminal-Bench 2.0 and Humanity's Last Exam, featuring improved coding skills, better code review and debugging, and a 1M token context window in beta. It outperformed GPT-5.2 by 144 Elo points on GDPval-AA economically valuable knowledge work tasks.

• State-of-the-art on Terminal-Bench 2.0
• Highest score on Humanity's Last Exam
• 1M token context window (beta)
• Improved agentic coding and computer use
• Outperforms GPT-5.2 on knowledge work tasks

Anthropic launched Labs as a team focused on incubating experimental products at the frontier of Claude's capabilities. The announcement highlighted a faster path from research previews into products that can scale reliably for customers.

• Incubates experimental frontier products
• Builds on Claude Code and MCP momentum
• Supports research previews and product scaling
• Focuses on agentic desktop experiences

Anthropic released Claude Opus 4.5, positioning it as a major step forward for coding, agents, and computer use. The release emphasized stronger real-world software engineering performance, better everyday productivity, and wider availability across Anthropic's products and partner platforms.

• Best-in-class coding and agent performance
• Strong real-world software engineering results
• Improved deep research and spreadsheet workflows
• Available on apps, API, and cloud platforms
• Lower pricing than prior Opus-tier releases

Anthropic released Claude Sonnet 4.5, describing it as the best coding model in the world and the strongest model for building complex agents. The release added major upgrades to Claude Code, the API, and the Claude app experience.

• Best coding model in the world
• Strongest model for complex agents
• Major Claude Code improvements
• Context editing and memory tool in the API
• Checkpoints and native VS Code extension

Anthropic released Claude 3.7 Sonnet, their first hybrid reasoning model that can produce both quick responses and extended thinking for complex tasks. It achieved state-of-the-art performance on SWE-bench Verified (62.3%) and featured enhanced agentic coding capabilities.

• First hybrid reasoning model
• Standard and extended thinking modes
• 62.3% on SWE-bench Verified
• State-of-the-art agentic coding
• 2x improvement in coding tasks

Anthropic released Claude Code, an agentic coding tool that works directly in the terminal. It can search and read code, edit files, run tests, and use command-line tools. Integrated with Claude 3.7 Sonnet, it achieved state-of-the-art performance on SWE-bench Verified.

• Terminal-integrated agentic coding
• Search, read, edit code
• Run tests and command-line tools
• State-of-the-art on SWE-bench
• Natural language project management

Anthropic released an upgraded Claude 3.5 Sonnet with 'computer use' capabilities, allowing the AI to directly interact with computers - moving the mouse, clicking buttons, and typing text. This enabled AI agents to perform complex multi-step tasks autonomously.

• Direct computer interaction capability
• Can move mouse and type
• Performs complex multi-step tasks
• Computer use API in public beta
• First frontier model with computer use

Anthropic released Claude 3.5 Sonnet, which outperformed Claude 3 Opus on many benchmarks while being faster and cheaper. It achieved 92% on HumanEval coding benchmark and demonstrated significant improvements in reasoning, coding, and following complex instructions.

• Outperformed Claude 3 Opus
• 92% on HumanEval coding benchmark
• 2x faster than Claude 3 Opus
• Better at following complex instructions
• Artifacts feature for creating interactive content

Anthropic released the Claude 3 family (Haiku, Sonnet, Opus), featuring near-human comprehension levels and vision capabilities. Claude 3 Opus set new benchmarks, outperforming GPT-4 on many tasks. All models featured improved speed, accuracy, and 200K context windows.

• Three models: Haiku, Sonnet, Opus
• Near-human comprehension on tasks
• Vision and multimodal capabilities
• 200K token context window
• Outperformed GPT-4 on many benchmarks

Anthropic released Claude 2, featuring 100K token context window, improved reasoning and coding capabilities, and strong safety measures. It achieved 71.2% on the Codex HumanEval coding benchmark and demonstrated Claude's focus on helpful, harmless, and honest AI.

• 100K token context window
• 71.2% on Codex HumanEval
• Improved reasoning and coding
• Safety via Constitutional AI
• Public API availability