India's Sarvam AI denies using subsidised GPUs to host Chinese open models

HimanshiET · x · 2026-11-19

After backlash over hosting Chinese models (DeepSeek V4.1 Flash, Kimi K3.1 Pro, GLM 5.3), India's Sarvam AI says its Token Factory inference platform runs entirely on privately procured GPUs, with its subsidised access to 4,096 NVIDIA H100s under the IndiaAI mission having ended in March 2025. The company says it has no commercial agreements with DeepSeek, Moonshot or Zhipu, and frames the platform as India's answer to OpenRouter; it raised $200M at a $1.5B valuation.

Original post →

More from Companies & People

Companies & People channel →