Anthropic, released Claude 4 (Sonnet 4 and Opus 4),Good at Coding
On May 22, 2025, Anthropic officially unveiled its next-generation Claude 4 family—Claude Opus 4 and Claude Sonnet 4—marking a significant leap in large-language model capabilities. The announcement, delivered at Anthropic’s inaugural developer conference, emphasizes Claude Opus 4’s prowess in sustained, long-horizon tasks—capable of “focused effort” over several hours—and positions Anthropic in direct competition with OpenAI, Google, and other frontier AI developers .
The Claude 4 series comprises two variants:
- Claude Sonnet 4, available to both free-tier and paid users, serves as a drop-in upgrade from Sonnet 3.7, offering improved coding, math, and instruction-following performance.
- Claude Opus 4, reserved for paying customers and API partners, represents the family’s flagship model, delivering world-class coding capabilities and advanced reasoning. It is accessible via Anthropic’s API on Amazon Bedrock and Google Vertex AI at $15/$75 per million tokens for Opus 4 (input/output) and $3/$15 per million tokens for Sonnet 4.
Benchmarked performance underscores Claude 4’s technical advances: Opus 4 claims a leading SWE-bench score of 72.5% and Terminal-bench of 43.2%, while maintaining “amnesia-resistant” memory over seven-hour coding sessions—demonstrated in real-world testing at Rakuten. Early reports also suggest that Claude 4 outpaces OpenAI’s comparable models in response quality and speed, potentially signaling Anthropic’s breakthrough over previous architectural limitations .
Given the heightened capabilities, Anthropic has activated its Responsible Scaling Policy (RSP), applying AI Safety Level 3 (ASL-3) safeguards to Opus 4. These include enhanced prompt-classification filters, anti-jailbreak defenses, and a vulnerability-bounty program—measures prompted by internal tests indicating Opus 4’s ability to guide novice users in creating harmful biothreats .
Anthropic’s launch arrives amid an intensifying AI arms race. Microsoft recently introduced a new coding agent alongside Elon Musk’s Grok on its platforms, while Google continues to integrate Gemini into search and productivity tools. At the same time, OpenAI has pursued hardware innovations via its acquisition of Jony Ive’s startup. Anthropic’s Claude 4 push seeks to maintain momentum, leveraging its $2 billion-plus annual revenue—projected to grow to $12 billion by 2027—and a fresh $2.5 billion credit facility alongside investments from Amazon and other backers .
As enterprises and developers begin migrating to Claude 4, early adopters are evaluating its integration into coding workflows, research applications, and autonomous agent frameworks. With its combination of raw performance, extended memory, and comprehensive safety measures, Claude 4 is poised to redefine expectations for AI collaboration—and to shape the emerging standards for responsible frontier‐model deployment.
Getting Started
CometAPI provides a unified REST interface that aggregates hundreds of AI models—including Claude family—under a consistent endpoint, with built-in API-key management, usage quotas, and billing dashboards. Instead of juggling multiple vendor URLs and credentials.
Developers can access Claude Sonnet 4 API (model: claude-sonnet-4-20250514 ; claude-sonnet-4-20250514-thinking) and Claude Opus 4 API (model: claude-opus-4-20250514; claude-opus-4-20250514-thinking)etc through CometAPI. To begin, explore the model’s capabilities in the Playground and consult the API guide for detailed instructions. Before accessing, please make sure you have logged in to CometAPI and obtained the API key.
