Judging from the questions raised by netizens to DeepSeek, the model’s answer is that DeepSeek R2 is expected to be released between August 15th and August 30th, 2025.

In response, a person close to DeepSeek stated that the news was untrue and confirmed that there were no plans to release DeepSeek R2 within August.

No matter when DeepSeek R2 is released, judging from their actual actions, the relevant version has indeed been updated and tested internally. However, it is still unknown at what point it will be launched.

It is reported that although the R&D team has been engaged in high-intensity development for several months, the company’s CEO, Liang Wenfeng, is still not satisfied with the performance of the new model as to why it has not been released yet.

In addition, DeepSeek may be facing the predicament of a shortage of computing power reserves. It is reported that due to the unexpected ban on the sale of NVIDIA’s H20 chip before, the computing power reserves of DeepQuest have been affected. The company is worried that once the new model is launched, the sudden surge in call volume in a short period of time may lead to a poor user experience due to insufficient computing power.

The post The company’s own large model claims that DeepSeek R2 will be Released in August appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

It is introduced that DeepSeek-Prover-V2-671B has a parameter count of 671 billion, uses a more efficient safetensors file format, and supports multiple calculation accuracies such as BF16, FP8, and F32, facilitating the faster and more resource-efficient training and deployment of the model.

In terms of model architecture, this model uses the DeepSeek-V3 architecture, adopts the MoE (Hybrid Expert) mode, and has 61 Transformer layers and 7,168-dimensional hidden layers.

It simultaneously supports ultra-long contexts, with the maximum position embedding reaching 163,840, enabling it to handle complex mathematical proofs. Moreover, it adopts FP8 quantization, which can reduce the model size and improve the reasoning efficiency through quantization technology.

Some netizens analyzed that this model is regarded as an upgraded version of Prover-V1.5, focusing on the proof of formal theorems, specifically used to solve mathematical problems, and is good at automatically proving theorems and complex calculations, similar to the self-playing style of AlphaGo in Go.

It is worth looking forward to how its performance test will go next.

The post New arrivals from DeepSeek! The new model DeepSeek-Prover-V2-671B was released as open source appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

Supported models are as follows:

Honor Tablet: Honor Tablet V9, Honor Tablet MagicPad2, Honor Tablet GT Pro

HONOR Mobile phone: HONOR Magic7/6/5 series, HONOR Magic V3/V2/V/Vs3/Vs2/Vs series, HONOR Magic V Flip, Honor 300/200 series, Honor GT.

Before using this function, please upgrade the mobile phone system to the latest version, and update the Honor Document App to the latest in the Honor App Market.

In February this year, Honor announced the official access to DeepSeek, becoming the first Android access to DeepSeek manufacturers.

After the launch of the DeepSeek-R1 Internet version of Honor YOYO agent, it not only has a strong deep thinking ability, but also can combine the Internet search function, in-depth search and integration of real-time network information, deep thinking and accurate answers to user questions.

At the same time, Honor YOYO will provide more efficient services in the aspects of dialogue and chat, text writing, language translation, programming assistance, literature summary, travel planning, etc., to improve users’ quality of life, work and learning efficiency.

The post Honor document officially connected to DeepSeek full blood version appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

DeepSeek-R2 delivers breakthroughs in several key areas, including better programming capabilities, multilingual reasoning capabilities, and greater accuracy at a lower cost. The corresponding sources believe that these features, if realized, could give it a significant advantage in the global AI race.

As of the Igeekphone press release, DeepSeek has not officially announced the specific date and technical details of R2, and the market earlier expected the DeepSeek-R2 model to be released in May.

The post The DeepSeek-R2 AI model will be released on March 17 appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.

At the infrastructure level, UniCube provides a pre-integrated intelligent computing cluster, which realizes the unified orchestration of network, storage and security management, bringing unprecedented convenience and efficiency to users.

At the application enable level, UniCube has built in the H3C AlStore smart asset platform, which provides free access to more than 1,000 selected open source datasets, models and images. At the same time, commercial resources can also be easily obtained by one-click payment, making the time for users to obtain AI assets such as hundreds of GB images and models reduced from days to hours, greatly accelerating the development process of AI applications.

In particular, UniCube has adopted a “turnkey” delivery model, ensuring rapid deployment and rollout of products. Its built-in Web front-end visual operation interface and standardized API interface greatly reduce the threshold of operation and reduce the workload of repeated development. Thanks to these innovative designs, application uptime has been significantly reduced from the traditional 1 week to less than 1 day, bringing significant time and cost benefits to users.

In terms of application scenarios, H3C large model all-in-one machine shows strong versatility and flexibility. It not only supports the distillation, fine-tuning, and reasoning of DeepSeek and other large models, but also seamlessly integrates these operations into workflow orchestration, providing users with a one-stop AI service experience. Whether it is industry data capitalization, model tuning building blocks, or free orchestration of AI application workflows, UniCube can easily deal with it and achieve efficient end-to-end delivery of industry AI applications.

The post Xinhua 3 DeepSeek all-in-one machine launched: equipped with DeepSeek V3 and R1 models appeared first on IGeeKphone China Phone, Tablet PC, VR, RC Drone News, Reviews.