DeepSeek v3.1 has emerged as a formidable competitor to OpenAI's gpt-oss-20b, showcasing superior performance in coding and creative writing. Released with minimal fanfare, DeepSeek’s hybrid architecture reportedly generates functional code and engaging narratives more effectively than OpenAI’s latest model, which has faced criticism for overthinking and providing poor output. In coding tests, DeepSeek produced a complex 2D maze game successfully, while OpenAI's model failed to deliver usable code under similar conditions. In a creative writing prompt, DeepSeek crafted a captivating story, contrasting with OpenAI's more intellectual and less engaging narrative. Furthermore, DeepSeek demonstrated better emotional intelligence in handling sensitive topics, providing nuanced responses that addressed underlying issues. Despite the current advantage of DeepSeek in execution and usability, OpenAI retains a lead in customization with specialized versions already developed by the community. The future may see improvements across both models as developers explore their potential.

Source 🔗