DeepSeek’s updated R1 AI model matches Google, Anthropic in coding benchmark

DeepSeek’s updated R1 AI model matches Google, Anthropic in coding benchmark

The updated reasoning model, released in May, performed well against leading US models in the real-time WebDev Arena tests.

How DeepSeek Rewrote the Transformer

How DeepSeek Rewrote the Transformer [MLA] How DeepSeek Rewrote the Transformer [MLA] Technical Notes 1. Note that DeepSeek-V2 paper claims a KV cache size reduction of 93.3%. They don’t exactly publish their methodology, but as far as I can tell it’s something likes this: start with Deepseek-v2 hyperparameters here: https://huggingface.co/deepseek-ai/DeepSeek-V2/blob/main/configuration_deepseek.py. num_hidden_layers=30, num_attention_heads=32, …

The future of AI in business: shifting from automation to autonomy

The year kicked off with an earthquake rattling the foundations of AI. Deepseek’s sudden arrival demonstrated there are multiple avenues in the advancement of super intelligence and drew the movers and shakers to weigh up software/algorithms versus banking on increases of model complexity or computational powers.

DeepSeek’s R1 upgrade takes on GPT-4 with some “rumoured” help from Gemini’s brain!

While China’s most ambitious open-source model may have been quietly fed by one of its Western rivals, if the product is an Open Source LLM better than GPT-4, does anyone really care? A couple of months ago we posted about DeepSeek training AI for pennies on the dollar and then another one about an entire [...] The post DeepSeek’s R1 upgrade takes on GPT-4 with some “rumoured” help from Gemini’s brain! appeared first on Sify.