E2E Networks544783New product launch

Launches 'Blaze', a pay-per-token inference service for open AI models, in India and globally

E2E Networks has launched a new product, 'Blaze' — a pay-per-token inference service for open AI models.

By ScoutQuest · Mon 2 Nov 2026, 4:49 PM IST · 2 min read

E2E Networks has launched a new product, Blaze, described as a pay-per-token inference service for open AI models.
Date of launch: 02-11-2026. The product caters to both the domestic and international market, and is launched globally.
Developers and enterprises get a single OpenAI-compatible endpoint to open models including NVIDIA Nemotron, Sarvam, DeepSeek, Qwen, GLM and Gemma.
Blaze is powered by NVIDIA GPU infrastructure and open-source components from NVIDIA DSX OS, and turns the company's India-hosted AI factory capacity into production inference services, with E2E managing model hosting and scaling.

Instead of renting a server for a fixed period, a customer pays only for the tokens (the small units of text an AI model reads and writes) that they actually use.
The company says Blaze is designed for inference-heavy products where latency and cost per token matter most, including voice, coding agents, document AI and OCR, image and video generation, and language applications.
The update states open weight models benefit from high KV cache reuse to reduce inference cost.

The company's AI factory capacity is expanding to about 14.6 MW.
This includes a 1,024-GPU NVIDIA Blackwell GPU cluster live since May 2026.
It also includes an additional 5.4 MW of Blackwell capacity including HGX B300 systems and other Blackwell GPUs.

Blaze and TIR, E2E's AI platform, are presented as one path from training to tokens: Blaze for API-based inference on open-weight models, and TIR for self-hosted inference and model pre-training and post-training on NVIDIA Blackwell GPU clusters through managed Kubernetes.
Blaze uses NVIDIA Dynamo for distributed model serving, NVIDIA Grove and KAI Scheduler to coordinate workloads across the Kubernetes fleet, and NVIDIA BlueField DPUs, NVIDIA DOCA and NVIDIA Infra Controller for hardware-level isolation between customers.
The update notes that such isolation matters for banks, healthcare providers and public-sector bodies with strict data rules.

Tarun Dua, Founder and Managing Director, said there is customer demand for frontier-class open weight models running on an NVIDIA Cloud Partner with NVIDIA reference design clusters as Sovereign Compute, and that Blaze provides this on E2E Cloud in India.
Vishal Dhupar, Managing Director for South Asia at NVIDIA, said access to AI computing and inference services is essential to advancing India's AI ecosystem and that NVIDIA's platform helps cloud providers like E2E Networks make AI more accessible.

The intimation is about a product being made available; customers are billed per token used.
The launch sits alongside an expansion of the company's AI factory capacity, and the update links the two — Blaze runs on that India-hosted capacity.

Also from E2E Networks

26 Oct 2026

26 Oct 2026

21 Oct 2026