Today, iFlytek Open Platform officially launched its 807th AI capability—the full range of DeepSeek models, provided by Hangzhou DeepSeek.
The full range of DeepSeek models includes two versions: DeepSeek-V3 and DeepSeek-R1, both of which support public cloud API calls and one-click deployment of dedicated models. Notably, the iFlytek Open Platform has also introduced the DeepSeek large model fine-tuning service, allowing developers to deeply customize a "smarter and more suitable" dedicated large model based on business scenario data. This service will undoubtedly greatly enhance developers' work efficiency and the applicability of the models.
In terms of performance, the DeepSeek R1 distilled model fine-tuning service has achieved a significant increase in inference speed through distillation optimization based on high-quality open-source small models, reportedly exceeding three times. At the same time, deployment costs have dropped by 80%, making the model more lightweight while maintaining strong logical reasoning capabilities, allowing it to run efficiently in various environments.
To ensure users can use the full range of DeepSeek models stably and smoothly, the iFlytek Open Platform relies on a hybrid cloud architecture, providing strong load balancing and automatic failover capabilities. Additionally, the platform has established a commercial production-level SLA service guarantee of 99.97%, ensuring 24/7 service support and response, giving users peace of mind.
To assist enterprises in low-cost trial and high-precision implementation, the iFlytek Open Platform also provides full-link tool support and a developer ecosystem. From data preprocessing to private deployment, the platform will accompany developers throughout the process to ensure the project can progress smoothly. Meanwhile, the iFlytek Star MaaS platform will continue to support developers in achieving intelligent upgrades at lower costs.
To celebrate the launch of the full range of DeepSeek models, the iFlytek Open Platform has also launched a limited-time free activity. From now until March 10, the DeepSeek inference API will be free for all users. Additionally, users can enjoy the convenience of "one-click deployment" for dedicated models, paying based on instance runtime, making it flexible and efficient.