Summary: SenseTime has launched the enterprise version of its DeepSeek large model platform, while also expanding the capabilities of its free version, creating a dual-engine approach to cater to a wider range of users.
Artificial intelligence firm SenseTime is doubling down on its DeepSeek large model platform, unveiling a new enterprise edition while simultaneously expanding the capabilities of its free version. This dual-engine strategy aims to provide both robust, professional-grade support and widespread accessibility.
Back in February, SenseTime’s SenseCore AI platform officially launched the DeepSeek-R1 and DeepSeek-V3 models, offering a generous 10 million token allowance for the free version over three months. This initial offering was met with enthusiastic responses from users. Now, after a series of optimizations, SenseTime is officially launching the DeepSeek-R1-Enterprise edition, designed to meet the pressing demands of various industries and enhance the user experience.
DeepSeek-R1-Enterprise: Power and Stability for Business
The enterprise version of DeepSeek-R1 promises high performance and stability, serving as a solid foundation for business applications. Key features include:
- Leading Inference Performance: Optimized with SenseTime’s proprietary LightLLM, featuring Prefill/Decode enhancements for industry-leading inference speeds.
- Parallel Computing Strategies: Leverages expert, pipeline, tensor, and data parallelism techniques to maximize hardware utilization and inference efficiency.
- Comprehensive Monitoring System: Tracks over 100 metrics, including first-packet latency, QPS (Queries Per Second), and throughput, providing a one-stop monitoring and alerting system.
Furthermore, SenseTime offers expert services to enterprise users, providing end-to-end solutions, including customized lightweight model development and scenario-based reinforcement learning to enhance base models for specific applications. Distillation to smaller, efficient models is also supported.
Free Version Gets a Boost: Localization and Expanded Compute
In addition to the enterprise launch, SenseTime has also enhanced the free version of the DeepSeek platform. These improvements include:
- Domestic Hardware Adaption: Completed the localization adaptation of the free version, ensuring compatibility with domestic hardware.
- Compute Capacity Expansion: Increased the available compute resources for the free version.
These upgrades to the free version, coupled with the launch of the enterprise edition, position SenseTime to offer a comprehensive suite of large model services, catering to both individual users and large organizations. This dual-engine approach, combining professional-grade support with widespread accessibility, could be a significant step forward in democratizing access to powerful AI tools.
References:
- SenseTime Official Website: https://console.sensecore.cn/aistudio
Views: 0