NVIDIA, Blackwell GPU'lar için geliştirdiği yazılım optimizasyonları sayesinde DeepSeek v4 modelinde token başına maliyeti 5 kat azalttı. Bu gelişme, yapay zeka çıkarım maliyetlerinde yeni bir standart belirleyerek şirketlerin operasyonel giderlerini düşürüyor.
Hong Kong Generative AI Research and Development Centre (HKGAI) launched HKGAI-V3, a new large language model based on DeepSeek V4. It offers tenfold efficiency in token compression and a hundredfold increase in agent runtime, running on domestic chips for potential export of Chinese AI.
A new open-source script swaps Claude Code's expensive Anthropic backend for DeepSeek V4 Pro, OpenRouter, or Fireworks AI—keeping the agent loop, slashing the bill.
NIST's CAISI evaluated DeepSeek V4 Pro using private benchmarks and a cost-comparison filter that excluded every US model except GPT-5.4 mini. Critics call the methodology convenient.
The Hong Kong Generative AI Research and Development Centre (HKGAI) is set to release HKGAI-V3, a sovereign AI model built on DeepSeek architecture optimized for domestic Chinese chips like Huawei's Ascend 910C.
DeepSeek's launch of its V4 AI model could trigger a broad reassessment of stocks across the AI industry chain from chipmakers to LLM developers, analysts say. The breakthrough model is described as the most powerful open-source platform capable of challenging US rivals OpenAI and Anthropic, delivering top-tier coding performance and significant advantages in reasoning and agentic tasks. AI chipmakers including Cambricon Technologies and Moore Threads Technology stand to benefit from increased demand for high-performance chips produced in mainland China. China's AI chip market could rise to 1.34 trillion yuan by 2029 from 142.5 billion yuan in 2024, representing 54% annual compound growth.
DeepSeek V4 yapay zeka modelinin piyasaya sürülmesiyle Huawei Ascend 950 çiplerine olan küresel ve yerel talep hızla yükseldi. ByteDance, Tencent ve Alibaba gibi Çin'in önde gelen teknoloji devleri, yeni nesil yapay zeka iş yüklerini yönetebilmek amacıyla Huawei ile stratejik görüşmelere başladı. Ascend 950 işlemci ailesi, DeepSeek V4-Flash ve V4-Pro modelleriyle tam uyumlu çalışacak şekilde optimize edildi.
DeepSeek is touting its ultra-low charges for models unveiled last week, stoking price competition as China challenges the United States in artificial intelligence (AI). Developers are being offered 75 per cent discounts on the newly released DeepSeek V4 Pro. Fees for input cache hits across the Hangzhou-based company’s AI platforms have also been slashed to 10 per cent of the original price. That provides a cost saving for users who frequently make similar or repeat requests. The V4 Pro and the...
OpenClaw has adopted DeepSeek's V4 Flash as the default model for its AI agent, expanding its catalogue to include V4 Flash and V4 Pro models. The update also integrated Google Meet and optimized multi-step task consistency. DeepSeek's V4 models support mainstream agent tools including Anthropic's Claude Code, OpenClaw, and Tencent's CodeBuddy. The release gained global attention due to Huawei's announcement of full support from its Ascend chips and supernode systems.
DeepSeek's V4 Pro model ranked second globally among open-source AI models, trailing Moonshot AI's Kimi K2.6, according to benchmark firm Artificial Analysis. While V4 Pro showed improvement over its predecessor V3.2, it still lagged behind top competitors, highlighting challenges facing Chinese AI firms racing to close the gap with the US amid intensifying competition and computing power constraints.
The Chinese lab that shook Wall Street just dropped its biggest, most efficient model yet, hours after OpenAI launched GPT-5.5.
Das chinesische Start-up DeepSeek hat eine neue KI-Generation veröffentlicht. Das Modell setzt auf eine neue Architektur und bietet ein größeres Kontextfenster.
DeepSeek, V4 serisini piyasaya sürdü. V4-Pro 1.6 trilyon parametreyle açık kaynaklı modeller arasında matematik ve kodlamada lider performans iddiasında. Google'ın Gemini 3.1-Pro'sunun ardından ikinci sırada. Huawei Ascend çiplerine özel optimizasyon, ABD kısıtlamalarına karşı Çin'in çip bağımsızlığına önemli bir adım olarak değerlendiriliyor. Şirket 20 milyar doları aşan değerlemeyle fon arayışında.
