DeepSeek introduces the fresh Prover-V2 model. Chinese AI accelerates in the race against the West

Speculations around DeepSeek's Chinese start-up, specializing in artificial intelligence, scope zenith. erstwhile the company presented a fresh model this week, the network was buzzing. However, it turned out that this is not a long-awaited R2 model, but only an improved version of the existing solution.

Prover-V2 – next step, but not breakthrough

On Wednesday DeepSeek published an update of his model Prover, designated as Prover-V2. The fresh version offers improved possibilities for solving mathematical evidence, but is inactive based on architecture DeepSeek-V3, presented last December.

Though it's not a breakthrough. R2, which manufacture has been waiting for for weeks, netizens speculate that the update may be the reason for the future launch of the fresh generation model.

AI Giants Race: Alibaba, Meta and Xiaomi in the game

DeepSeek update coincided with the announcement New generation models Qwen Alibaba. The company boasted that Qwen3 exceeds DeepSeek-R1 In selected benchmarkes, which became a benchmark for Chinese AI models.

Analysts indicate that the latest Alibaby models importantly reduce the gap between Chinese and American solutions. However, as experts point out, these comparisons may be ambiguous – they depend on the comparative methodologies and user priorities adopted.

In this context, words are interesting Mark Zuckerbergwhich in the podcast with Dwarkesh Patel acknowledged that Meta stopped optimizing its models Llama for benchmarking, focusing on actual generic AI applications. The company has just released fresh API for Llama modelsallowing developers to integrate them.

Is that a convenient explanation for the fact that Llama lost her leadership position among open-source AI models for Qwen? Zuckerberg admits, however, that the improvement of the ecosystem of open solutions is good news for the full industry.

DeepSeek inspires: Xiaomi enters the game and Musk announces the fresh Grok

DeepSeek's success made the next tech giants decide to open their models. This week Xiaomi joined the race, presenting its first open-source applicant model.

Outside China Elon Musk He reacted quickly, announcing improved version of Grok, developed by its start-up xAI. Unlike Chinese solutions, the latest iterations Grok are not open-source.

The War of Talents: DeepSeek, Ant Group and Tsinghua search the best

DeepSeek affects Chinese AI marketplace not only through technology, but besides through recruitment policy. The company employs promising developers Right after college., frequently graduates of leading universities specified as Tsinghua and Beijing University.

In consequence Ant Group, the Fintech branch of Alibaby, praised his investigation team, trying to attract specialists under conditions of limited talent supply.

Meanwhile Tsinghua University strengthens its AI department by hiring Alexa LambaFormer Microsoft researcher. Lamb confirmed in an interview with South China Morning Postto join the recently created College of AIwho presently recruits candidates for his program.

Waiting for the Revolution

Come on. Prover-V2 This is simply a step forward, the technology community is inactive looking DeepSeek-R2 – a model that can bring a real breakthrough. Meanwhile, the AI marketplace accelerates and China strengthens its position in the global race, competing with Western giants.

Source: South China Morning Post – Inside China Tech, 03.05.201025

Leszek B. Glass

Email: [email protected]

© www.chiny24.com