https://it-news.uk/tags/deepseek/ Recent content in Deepseek on IT News Hugo en-us Wed, 08 Jul 2026 00:00:00 +0800 https://it-news.uk/posts/deepseek-designing-own-ai-chip-inference/ Wed, 08 Jul 2026 00:00:00 +0800 https://it-news.uk/posts/deepseek-designing-own-ai-chip-inference/ <p>DeepSeek, the Chinese AI startup that has made waves with its open-weight models and aggressive pricing, is now quietly building its own chips. If the project pays off, it could fundamentally rewire who DeepSeek depends on — and who competes with whom in China&rsquo;s AI hardware market.</p> <p>Three people familiar with the matter told Reuters that DeepSeek is developing a custom AI chip designed specifically for inference — the phase where a trained model actually processes new data and generates output, as opposed to the training phase that gobbles up GPU clusters for weeks. The project is about a year old and still in its early stages, the sources said.</p> <p style="text-align: center"><img src="https://it-news.uk/images/itnews-973810-0.jpg" alt="" style="text-align: center"></p> https://it-news.uk/posts/nvidia-deepseek-v4-blackwell-inference-80-percent-cost-cut/ Wed, 01 Jul 2026 00:00:00 +0800 https://it-news.uk/posts/nvidia-deepseek-v4-blackwell-inference-80-percent-cost-cut/ <p>Running a large language model in production is expensive. Every API call burns compute, and the bills pile up fast. NVIDIA just published the numbers on how much that equation has changed for one of the most popular models on the market.</p> <p>The company announced Tuesday that its Blackwell platform, running DeepSeek V4 with a fully optimized inference stack, now delivers per-token costs as low as one-fifth of what they were a month ago. Token throughput per GPU has improved by as much as 20x.</p> https://it-news.uk/posts/deepseek-v4-july-launch-peak-pricing/ Mon, 29 Jun 2026 00:00:00 +0800 https://it-news.uk/posts/deepseek-v4-july-launch-peak-pricing/ <p>DeepSeek announced Monday that the production version of DeepSeek V4 will ship in mid-July, bringing performance improvements and a pricing model that charges developers more during peak hours.</p> <p>The peak/valley system doubles API costs during two daily windows — 9:00 AM to 12:00 PM and 2:00 PM to 6:00 PM — while keeping rates at current levels during off-peak hours. It is a supply-and-demand mechanism designed to smooth traffic spikes and improve service stability, not a blanket price hike.</p> https://it-news.uk/posts/deepseek-dspark-open-source-inference-acceleration/ Sat, 27 Jun 2026 00:00:00 +0800 https://it-news.uk/posts/deepseek-dspark-open-source-inference-acceleration/ <p>There&rsquo;s a quiet tension inside every AI chatbot: the model generates text one token at a time, each step requiring a full forward pass through billions of parameters. The longer the output, the longer the wait. DeepSeek, working with researchers at Peking University, just open-sourced a framework called DSpark that tackles this bottleneck head-on, and the numbers are striking.</p> https://it-news.uk/posts/deepseek-hiring-chinese-proficiency-response/ Tue, 23 Jun 2026 10:08:00 +0800 https://it-news.uk/posts/deepseek-hiring-chinese-proficiency-response/ <p>DeepSeek Harness team lead Cui Tianyi has pushed back against online accusations that the AI company refuses to hire foreigners, drawing a comparison to how American tech firms operate. In a social media post on June 21, Cui explained that the requirement is straightforward: employees must be able to work in Chinese — just as their US counterparts expect English fluency — and there is no policy barring foreign nationals from joining the company.</p>