- Nvidia integrates DeepSeek-R1 as a NIM microservice
- AWS helps DeepSeek-R1 with a give attention to scalable and cost-efficient AI deployment
- Microsoft additionally has future native deployment plans for DeepSeek
Having taken the AI world by storm in latest weeks, DeepSeek has now made important strides in increasing the accessibility of its superior reasoning fashions.
The corporate has introduced its flagship DeepSeek R1 mannequin is now accessible on a number of platforms, together with Nvidia, AWS, and GitHub.
DeepSeek’s open supply nature permits builders to construct fashions primarily based on its structure, and, at press time, there are 3,374 DeepSeek-based fashions accessible collaborative AI-model growth platform Hugging Face.
Nvidia, AWS, Github & Azure now gives DeepSeek
On AWS, DeepSeek-R1 fashions at the moment are accessible by means of Amazon Bedrock which simplifies API integration and Amazon SageMaker which permits superior customization and coaching, supported by AWS Trainium and Inferentia for optimized value effectivity.
AWS additionally gives DeepSeek-R1-Distill, a lighter model, by means of Amazon Bedrock Customized Mannequin Import. This serverless deployment simplifies infrastructure administration whereas sustaining scalability.
Nvidia has additionally built-in DeepSeek-R1 as a NIM microservice, leveraging its Hopper structure and FP8 Transformer Engine acceleration to ship real-time, high-quality responses.
The mannequin, which options 671 billion parameters and a 128,000-token context size, makes use of test-time scaling for improved accuracy.
It additionally advantages from Nvidia’s Hopper structure, utilizing FP8 Transformer Engine acceleration and NVLink connectivity. Operating on an HGX H200 system, DeepSeek-R1 can generate as much as 3,872 tokens per second.
Microsoft’s Azure AI Foundry and GitHub have additional expanded DeepSeek’s attain, providing builders a safe and scalable platform to combine AI into their workflows.
Microsoft has additionally carried out in depth security measures, together with content material filtering and automatic assessments. The corporate claims that it plans to supply distilled variations of DeepSeek-R1 for native deployment on Copilot+ PCs sooner or later.
DeepSeek-R1 took the world by storm by providing a strong, cost-efficient AI mannequin with superior reasoning capabilities and has dethroned widespread AI fashions like ChatGPT.
R1 was reportedly skilled for simply $6 million, with its most superior variations being about 95% cheaper to coach than comparable fashions from NVIDIA and Microsoft.