Listed AI cloud E2E Networks launches Blaze pay-per-token inference for open models in India

E2E Networks Ltd, the NSE- and BSE-listed Indian AI cloud company, launched Blaze on November 2, 2026, as a pay-per-token inference service for open AI models, according to its Regulation 30 stock-exchange filing and reports by Analytics India Magazine, Dataquest and Elets CIO. Customers reach models through a single OpenAI-compatible API while E2E hosts and scales the models on NVIDIA GPU infrastructure in India, and they pay only for the tokens they use.

Supported open-weight families named in the launch materials include NVIDIA Nemotron, Sarvam, DeepSeek, Qwen, GLM and Gemma. E2E said Blaze targets inference-heavy products such as voice, coding agents, document AI and OCR, image and video generation, and language applications. The product page says the service runs in Indian data centres and is billed in INR, with no need for customers to reserve GPU capacity themselves.

Under the hood, E2E said Blaze uses open-source components from NVIDIA DSX OS. NVIDIA Dynamo handles distributed serving with KV cache-aware routing and separate prefill and decode stages. NVIDIA Grove and the KAI Scheduler coordinate jobs across E2E's Kubernetes fleet, and BlueField DPUs with DOCA provide hardware-level isolation aimed at banks, healthcare and public-sector buyers. Founder and Managing Director Tarun Dua said customers want frontier-class open-weight models on sovereign compute in India.

Blaze sits beside TIR, E2E's existing platform for self-hosted inference and model training on managed Kubernetes. The company said AI factory capacity is expanding toward about 14.6 MW, including a 1,024-GPU NVIDIA Blackwell cluster live since May 2026 and a further 5.4 MW of Blackwell capacity that includes HGX B300 systems. Detailed per-token prices were not published in the launch materials reviewed by Dataquest.