https://it-news.uk/tags/inference/ Recent content in Inference on IT News Hugo en-us Wed, 01 Jul 2026 00:00:00 +0800 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/openai-inference-cost-halved-system-optimization/ Tue, 30 Jun 2026 00:00:00 +0800 https://it-news.uk/posts/openai-inference-cost-halved-system-optimization/ <p>There&rsquo;s a number inside every AI company that determines more than most people realize: the cost of running a single inference. For OpenAI, that number just got a lot smaller.</p> <p>IT-NEWS has learned, via a report from The Information, that OpenAI engineers have quietly pulled off a significant cost reduction — cutting the inference cost of their AI models by more than 50% through a series of system-level optimizations.</p> https://it-news.uk/posts/xiaomi-mimo-ultraspeed-trial-extended/ Tue, 23 Jun 2026 20:16:00 +0800 https://it-news.uk/posts/xiaomi-mimo-ultraspeed-trial-extended/ <p>Xiaomi announced today that it is extending the limited trial window for MiMo-V2.5-Pro-UltraSpeed, the high-performance inference mode that delivers up to 1,000 tokens per second. In a notification published on June 23, the company said the decision follows an unexpectedly massive wave of interest from the developer community.</p>