E2E Networks launched Blaze on November 2, introducing a pay-per-token service for running open artificial intelligence models on computing infrastructure hosted in India.
The company’s exchange-filed announcement lists NVIDIA Nemotron, Sarvam, DeepSeek, Qwen, GLM and Gemma among the model families supported through a single OpenAI-compatible application programming interface. E2E manages hosting and scaling while customers are charged for tokens used.
Blaze uses NVIDIA GPU infrastructure and software components including NVIDIA Dynamo. Intended applications include coding assistants, voice services, document processing and optical character recognition. These descriptions explain the product’s proposed uses, rather than establish independently tested performance.
The service is offered to domestic and international markets. It sits alongside E2E’s TIR platform, which supports model training, customisation and self-hosted deployment.
For developers, the practical distinction is access to managed model endpoints without arranging the underlying serving infrastructure themselves. The launch filing does not specify token prices or publish comparative latency benchmarks, leaving those details important for evaluating a deployment.
Archive illustration of data-centre server cabinets; not E2E’s facilities or Blaze hardware. Photo: Data Center Servers by Cory M. Grenier, CC BY-SA 2.0. Display may be resized or cropped.
