© IE Online Media Services Pvt Ltd
Tags:
Journalism of Courage
int(2)
US-based AI company Anthropic has introduced its latest AI model—Claude 3.7 Sonnet. The latest offering from Anthropic is hailed as a game changer for developers. It is essentially an iteration that builds on the foundation of Claude 3.6 Sonnet and Claude 3.5 Sonnet. The new model comes with an array of improvements aimed at developers. These new features enhance reasoning transparency and control over computational resources.
Anthropic has also introduced Claude Code, a research preview that offers some serious competition to existing AI-powered Integrated Development Environments (IDEs), most of which need paid subscriptions. Claude Code can be accessed directly through Claude, making it a lucrative tool for developers.
The two standout features of the new model are a transparent chain of thought and an extended thinking mode. With the transparent chain of thought, Claude 3.7 Sonnet is able to display raw, step-by-step reasoning abilities. Unlike some other AI models, especially from OpenAI, Claude Sonnet 3.7 offers full transparency allowing users to see how the model arrives at its conclusions. This can be beneficial for developers who want to see insights into the decision-making process.
On the other hand, the Extended Thinking Mode allows developers to dictate how long the model should process a problem. Developers here get the freedom to set a thinking budget meaning if a question is complex, they can define how many tokens the model can use e.g. 10,000 or 20,000 tokens. This level of control can likely enhance performance, keep costs in check, and also prevent overthinking on simpler tasks.
While Anthropic models have always been known for their superior coding abilities, Claude 3.7 Sonnet takes that a notch even further with the Claude Code. It is essentially a built-in code editor from Anthropic. With Claude Code, the AI company seems to be expanding into the arena of AI-assisted development tools, much like OpenAI.
Based on the benchmark results shared by Anthropic, the new model seems to have done exceptionally well. On the SWE-Bench (GitHub issue-solving test), the model scored 62.3 per cent and with extended thinking it reached 70.3 per cent. On agentic and coding use cases, it showed higher scores compared to OpenAI’s GPT-4o. Similarly, on Multi-Lingual Understanding (MLU), its performance improved from 83.2 per cent to 86 per cent with extended thinking enabled. The model got an accuracy of 96 per cent on Math 500, and on AI and ML tests, it got 61 per cent.
With groundbreaking features, the Claude 3.7 Sonnet comes with a higher price tag. The new model has a 200,000-token context window and a price that is higher than its OpenAI counterparts. Claude 3.7 Sonnet is priced at $3 per million tokens and $15 million output tokens, while GPT-4o Mini is at $1.10 per million input tokens and $4.40 per million output tokens. If Claude 3.7 Sonnet aims to directly compete with GPT-4o mini, then its pricing is steep; however, if it outperforms the model by leaps, then the price may be justifiable.
Claude 3.7 Sonnet is now available on Anthropic’s platform, offering free access with some limitations.
