NVIDIA just dropped a thermonuclear benchmark bomb—and Wall Street's algo-traders are already mispricing it.The GB200 NVL72 isn't an upgrade. It's an extinction-level event for legacy AI hardware, pai...

<p><p>NVIDIA just dropped a thermonuclear benchmark bomb—and Wall Street&#39;s algo-traders are already mispricing it.</p><p>The GB200 NVL72 isn&#39;t an upgrade. It&#39;s an extinction-level event for legacy AI hardware, pairing with OpenAI&#39;s GPT-OSS models to rewrite the rules of machine learning throughput.</p><p>72-node liquid-cooled beast mode</p><p>Forget incremental gains. This black slab of silicon delivers exponential leaps in training efficiency while drawing less power than a hedge fund&#39;s weekend cocaine budget.</p><p>Open-source arms race escalates</p><p>GPT-OSS models now chew through unstructured data like a Bloomberg terminal on margin calls—except with actual risk management. The combo creates perfect infrastructure for next-gen decentralized AI networks.</p><p>Will institutions finally wake up? Or keep pretending CUDA cores are &#39;just a fad&#39; while retail builds the future?</p></p><img width="600" height="315" src="https://image.blockchain.news:443/features/D8E08E86F8EDBDDCD68414CF49BDD8B1401B11A69515DFF98E6B2B03EE9CF9D7.jpg" alt="NVIDIA Boosts AI Performance with GB200 NVL72 and OpenAI gpt-oss Models"></p>

<p>NVIDIA, in collaboration with OpenAI, has announced significant advancements in AI performance, leveraging the power of the <a href="https://www.btcc.com/en-US/academy/financial-investment/nvidia-nvda-stock-price-prediction-202320252030-is-nvda-a-good-investment" title="Nvidia" rel="nofollow">Nvidia</a> GB200 NVL72 system. The recent launch of the OpenAI gpt-oss-20b and gpt-oss-120b models promises to deliver up to 1.5 million tokens per second (TPS), marking a substantial leap in AI processing capabilities, according to NVIDIA.</p>
<h2>Enhanced AI Capabilities</h2>
<p>The gpt-oss models, known for their text-reasoning capabilities, are built using the mixture of experts (MoE) architecture with SwigGLU activations. These models utilize RoPE for attention layers, supporting a 128k context length, and are optimized for NVIDIA&#39;s Blackwell architecture. They are released in FP4 precision, compatible with an 80 GB data center GPU, and optimized for NVIDIA&#39;s advanced hardware.</p>
<h2>Collaborative Developments</h2>
<p>NVIDIA&#39;s collaboration with OpenAI extends to various open-source frameworks, including Hugging Face Transformers and NVIDIA TensorRT-LLM, to enhance model performance and developer accessibility. The gpt-oss-120b model, in particular, required extensive training, amounting to over 2.1 million GPU hours.</p>
<h2>Technical Specifications</h2>
<p>The gpt-oss-20b and gpt-oss-120b models feature a range of specifications to cater to diverse AI needs. These include varying transformer block counts, total parameters, and expert configurations, designed to optimize inference performance on NVIDIA&#39;s platforms.</p>
<h2>Deployment Options</h2>
<p>NVIDIA offers multiple deployment options for developers, including the use of vLLM and TensorRT-LLM for server setup and performance optimization. The GB200 NVL72 system is designed to handle high throughput, accommodating up to 50,000 concurrent users efficiently.</p>
<h2>Future Prospects</h2>
<p>With the introduction of these advanced models, NVIDIA aims to support a broad spectrum of AI applications from cloud to edge. Their efforts to integrate gpt-oss models across various platforms highlight a commitment to enhancing AI infrastructure and developer experience.</p>
<p>For more details on the deployment and capabilities of these models, visit the NVIDIA blog.</p>
Image source: Shutterstock
<ul>
                                    <li>nvidia</li>
                                    <li>openai</li>
                                    <li>ai models</li>
                                    <li>gb200 nvl72</li>

                            </ul>

Followin

NVIDIA’s GB200 NVL72 &amp; OpenAI’s GPT-OSS Models Unleash AI Performance Frenzy

NVIDIA刚刚投放了一颗热核性能基准测试炸弹——华尔街的ALGO交易员已经对其定价不当。<GB200 NVL72>不是一次升级。这是对传统AI硬件的灭绝级事件，派...

<p><p>NVIDIA刚刚投放了一颗热核基准测试炸弹——华尔街的算法交易员已经错误定价了它。</p><p>GB200 NVL72不是一次升级。这是对传统AI硬件的灭绝级事件，与OpenAI的GPT-OSS模型配对，重新定义机器学习吞吐量的规则。</p><p>72节点液冷野兽模式</p><p>忘记渐进式收益。这块黑色硅板在训练效率上实现指数级跃升，同时耗电量比对冲基金的周末可卡因预算还低。</p><p>开源军备竞赛升级</p><p>GPT-OSS模型现在像彭博终端处理保证金通话一样处理非结构化数据——只是具有实际的风险管理。这种组合为下一代去中心化AI网络创造了完美基础。</p><p>机构最终会醒悟吗？还是继续假装CUDA核心只是"一时的潮流"，而散户正在构建未来？</p></p><img width="600" height="315" src="https://image.blockchain.news:443/features/D8E08E86F8EDBDDCD68414CF49BDD8B1401B11A69515DFF98E6B2B03EE9CF9D7.jpg" alt="NVIDIA通过GB200 NVL72和OpenAI gpt-oss模型提升AI性能"></p>

<p>NVIDIA与OpenAI合作，宣布在AI性能方面取得重大进展，利用Nvidia GB200 NVL72系统的强大功能。最近推出的OpenAI gpt-oss-20b和gpt-oss-120b模型有望每秒处理多达150万个代币，根据NVIDIA的说法，这标志着AI处理能力的实质性飞跃。</p>
<h2>增强的AI能力</h2>
<p>以文本推理能力著称的gpt-oss模型，采用专家混合（MoE）架构和SwigGLU激活。这些模型使用RoPE进行注意力层，支持128k上下文长度，并针对NVIDIA的Blackwell架构进行了优化。它们以FP4精度发布，兼容80 GB数据中心GPU，并针对NVIDIA的先进硬件进行了优化。</p>
<h2>协作发展</h2>
<p>NVIDIA与OpenAI的合作延伸到各种开源框架，包括Hugging Face Transformers和NVIDIA TensorRT-LLM，以提高模型性能和开发者可访问性。特别是gpt-oss-120b模型，其训练耗时超过210万GPU小时。</p>
<h2>技术规格</h2>
<p>gpt-oss-20b和gpt-oss-120b模型具有多种规格，以满足不同的AI需求。这些包括不同的Transformer块数量、总参数和专家配置，旨在优化NVIDIA平台上的推理性能。</p>
<h2>部署选项</h2>
<p>NVIDIA为开发者提供多种部署选项，包括使用vLLM和TensorRT-LLM进行服务器设置和性能优化。GB200 NVL72系统旨在高效处理高吞吐量，可同时容纳多达5万个并发用户。</p>
<h2>未来前景</h2>
<p>通过推出这些先进模型，NVIDIA旨在支持从云端到边缘的广泛AI应用。他们致力于在各个平台上整合gpt-oss模型，凸显了增强AI基础设施和开发者体验的承诺。</p>
<p>欲了解这些模型的部署和功能的更多详情，请访问NVIDIA博客。</p>
图片来源：Shutterstock
<ul>
                                    <li>nvidia</li>
                                    <li>openai</li>
                                    <li>ai模型</li>
                                    <li>gb200 nvl72</li>

                            </ul>

NVIDIA 的 GB200 NVL72 和 OpenAI 的 GPT-OSS 模型释放 AI 性能狂潮

在加密世界，真正决定胜负的从来不是代码，而是流动性、权力与监管之间的博弈。BUSD 的兴衰，并不是一枚稳定币的血泪史，而是一场关于交易所如何试图“统一天下”的金融战争。
当 BUSD 被推上历史舞台，它承载的不只是合规叙事，更是币安试图用高压手段重塑稳定币秩序的野心——通过强制汇率、合并交易对，把竞争对手的深度直接吞进自己的账本。
2022 年的“自动转换”，是一场教科书级别的闪击战；而 2023...

币安霸权兴衰录：从BUSD“大一统”到$U“吞天下”的稳定币战争

来源：Galaxy Research；翻译：金色财经xz
比特币2025年的收官价位预计将与年初水平大体持平。
今年前十个月，加密货币市场经历了一轮真正的看涨浪潮。监管改革持续推进，ETF持续吸纳资金，链上活动也日趋活跃。比特币在10月6日创下126,080美元的历史新高。
然而，市场在狂热情绪后并未迎来期待的突破，取而代之的是轮动、重定价和再平衡。宏观预期的落空、投资叙事的转向、杠杆的爆仓以及巨...

Galaxy：2026年加密市场26大预测

原创 | Odaily 星球日报（@OdailyChina）
作者 | Asher（@Asher_ 0210）
12 月 17 日，预测市场平台 predict.fun 的创始人 dingaling 在 X 平台发文宣布，平台已正式开放，并向 BNB Chain 上的 Meme 交易者、Aster 参与者、及各预测市场活跃的用户空投了初始积分。
与 Polymarket、Kalshi 等传统预测市...