At its first developer conference on Thursday, Anthropic unveiled its latest artificial intelligence (AI) models, Claude 4. The San Francisco-based company presented two new models, Claude Opus 4 and Claude Sonnet 4, while showcasing enhanced functionalities such as Extended Thinking with tool capability. Claimed to be state-of-the-art (SOTA) in coding, tool utilization, and writing, Opus 4 is positioned as the company’s premier AI model. Furthermore, Claude Code is now available for general use, with beta extensions accessible in VS Code and JetBrains, and it is also included among the offerings available on GitHub.
Anthropic Unveils Claude 4 AI Models
In a recent announcement, Anthropic detailed the features of its new models and the updates being implemented across its chatbot and application programming interface (API). The company has emphasized coding capabilities and agentic functions in its latest large language models (LLMs).
Opus 4 and Sonnet 4 are both hybrid models designed with two operation modes: near-instant responses and Extended Thinking for more profound reasoning. According to Anthropic, Opus 4 is labeled “the best coding model in the world,” boasting scores of 72.5 percent on the SWE-Bench and 43.2 percent on the Terminal-Bench, both of which assess coding performance.
Performance of Claude 4 models on the SWE-Bench
Photo Credit: Anthropic
In comparison, Claude Sonnet 4 has also shown marked enhancements over its earlier version. The company reported an internal evaluation score of 72.7 percent on SWE-Bench for Sonnet 4, asserting that it strikes a better balance between performance and efficiency despite a lower performance in some areas than Opus 4.
Claude Opus 4 features advancements in long-term task awareness and improved memory. Anthropic has rectified a previous issue where models would take shortcuts to complete tasks. During Extended Thinking sessions, both models are capable of utilizing tools, allowing them to alternate between internal reasoning and seeking external information, such as web searching, enhancing their responses. Additional improvements include the ability to use tools concurrently and better prompt compliance.
Both Opus 4 and Sonnet 4 are currently available for Claude Pro, Max, Team, and Enterprise subscribers, with Sonnet 4 also accessible to free users. Developers can utilize these LLMs through the Anthropic API, as well as on platforms like Amazon Bedrock and Google Cloud’s Vertex AI. Anthropic confirmed that pricing would remain consistent with the previous generation.
For developers, the cost for Opus 4 is set at $15 (approximately Rs. 1,290) per million input tokens and $75 (about Rs. 6,440) per million output tokens. In contrast, Sonnet 4 will be priced at $3 (roughly Rs. 260) per million input tokens and $15 (around Rs. 1,290) for each million output tokens.
In addition to the new AI models, Anthropic announced various features and made Claude Code widely available. Initially introduced in February as a research preview, Claude Code is an agentic coding tool capable of handling a broad array of coding tasks. Beta extensions have now been rolled out for VS Code and JetBrains, while an SDK for Claude Code is also being released in beta on GitHub.
