Context:

DeepSeek, a Chinese AI startup, has released an update to its R1 reasoning model, intensifying competition with US AI firms like OpenAI. The update, launched on the developer platform Hugging Face, has not been officially announced but was shared within a WeChat group as a 'minor trial upgrade' available for testing. The updated model ranks just behind OpenAI’s o4 mini and o3 models on the LiveCodeBench leaderboard, surpassing xAI’s Grok 3 mini and Alibaba’s Qwen 3. Earlier this year, DeepSeek's R1 model challenged industry norms by outperforming US models at lower costs, causing significant declines in US tech stock prices. Following R1's success, other Chinese firms like Alibaba and Tencent have developed competing models, while US companies have adjusted their pricing strategies to maintain competitiveness.

Dive Deeper:

• DeepSeek, a Chinese AI company, has released an update to its R1 reasoning model, labeled R1-0528, on the developer platform Hugging Face, although no official public announcement has been made yet.

• The R1-0528 update was shared in a WeChat group as a 'minor trial upgrade', and users have been invited to begin testing it, though specific details and comparisons of the model have not been published.

• The LiveCodeBench leaderboard, a benchmark created by UC Berkeley, MIT, and Cornell researchers, ranked DeepSeek's updated R1 model just behind OpenAI's o4 mini and o3 models in code generation, while it outperformed xAI’s Grok 3 mini and Alibaba’s Qwen 3.

• Earlier in the year, the initial release of DeepSeek's R1 model surprised the industry by performing on par with or better than leading US AI models at a fraction of the cost, which significantly impacted US tech stock prices.

• Following the success of DeepSeek's R1 model, Chinese tech giants like Alibaba and Tencent have introduced AI models claiming superior performance, leading US companies like Google and OpenAI to adjust their pricing and release new models to stay competitive.

• DeepSeek has plans to release a successor model, R2, initially scheduled for May, according to sources cited by Reuters, and has also upgraded its V3 large language model earlier in March.

• The competition in the AI industry has prompted significant shifts in strategies, with companies like Google launching discounted access tiers and OpenAI releasing a less computing power-dependent o3 mini model.