Anthropic has launched two more Claude AI types: Claude Sonnet 4 and Claude Opus 4.
Both Claude versions feature web searches into their responses with prolonged thinking mode, now in alpha. Anthropic says both have improved memory capabilities ( saving information across sessions ) and demonstrate improved instruction-following and contextual reasoning. People can switch rapid-response or long-form logic types depending on job difficulty.
Sonnet 4 is today available for free on the Claude gate, while people on Claude Pro, Max, Team, and Enterprise plans may gain access to both types. Pricing for API support is the same as for previous Anthropic models: Opus is priced at$ 15 for input and$ 75 for output per million tokens, while Sonnet 4 is priced at$ 3 and$ 15, respectively.
Claude Opus 4 can freely adopt’ dozens of ways ‘
Anthropic says Claude Opus 4 outperforms competitors in AI coding benchmarks and is capable of autonomously completing complex tasks involving” thousands of steps”, even operating continuously for several hours. Anthropic tailored Claude Opus 4 for greater precision and better understanding of codebases than previous models.
SEE: Apple is rumored to be preparing a SDK for its on-device AI models.  ,
Claude Sonnet 4 brings greater coding skills for free
Claude Sonnet 4, now Anthropic’s default general-purpose model, also introduces some tweaks specifically to boost its coding prowess. Sonnet 4 has greater steerability and scored higher than its predecessor on the SWE-bench software engineering test.
Anthropic promotes Claude Code tool to general availability and adds options to the Anthropic API
Anthropic has opened the Claude Code tool up to the general public and added native integrations to VSCode and JetBrains. A Claude Code SDK, now on Git Hub, can be used to build AI agents.
Additionally, four new capabilities for the Anthropic API are now in beta: a code execution tool, MCP connector, Files API, and the ability to cache prompts for up to one hour.
