DeepSeek R1 has achieved a score of 57% on the Aider polyglot benchmark, ranking second behind o1, which scored 62%. Other competitors included Sonnet at 52% and DeepSeek Chat V3 at 48%. The leaderboard highlights the performance of these models in advanced reasoning and search capabilities. Users have noted that DeepSeek R1 excels in web searching, matching the performance of GPT-4o, and it features a 'Deep Research' capability that integrates search and reasoning, positioning it competitively against similar features from Gemini and Perplexity. Feedback from users suggests that DeepSeek R1 may have advantages over its competitors, particularly in accessing the web and handling complex queries, although some noted its tendency to produce unnecessary code outputs.
Now you can combine DeepSeek-R1 with search to think in the web space. https://t.co/TP82TGHpd7
BREAKING 🚨: It turns out that DeepSeek also has a "Deep Research" feature which allows you to combine search and reasoning options. This solution may work well for complex search queries to compete with similar features on Gemini and Perplexity 👀 https://t.co/oOOx6UrkwU https://t.co/vUJFo5f3qi
Looks like DeepSeek r1 can access the web. Advantage over o1. DeepSeek delivered hard. https://t.co/MMCck9WKdJ