0G launches 1.6T-parameter DeepSeek V4 Pro with TeeTLS routing to Alibaba Bailian

0G says DeepSeek V4 Pro is live on 0G Private Computer through Router Mode and Advanced Mode, using the same `app-sk-*` key flow and `router-api.0g.ai/v1` endpoint as 0GM-1.0. The 1.6T-parameter MoE model has 49B active parameters, a 1M-token context window, and 384K max output, with inference hosted by Alibaba Bailian while 0G adds TeeTLS-attested transport and onchain proof of the routing path. This is not enclave-bound inference like 0GM-1.0; the pitch is verifiable routing and enterprise-term access for teams that cannot send prompts through DeepSeek’s direct PRC API, priced at $2.10/M input and $4.20/M output in $0G.

TLDR by @Benthic

More coverage

• DeepSeek V4 Preview Release (api-docs.deepseek.com)
• deepseek-v4-pro Model by Deepseek-ai | NVIDIA NIM (build.nvidia.com)
• DeepSeek V4 Pro: Model Overview, Features & Performance Guide (deepinfra.com)
• deepseek-ai/DeepSeek-V4-Pro - Hugging Face (huggingface.co)
• DeepSeek V4: the most expected open-source model ever released ... (lambda.ai)

Explore the topic

• Launch