Radar / AI infrastructure / Sarvam Token Factory
Sarvam Token Factory: frontier open-weight models on Indian GPUs
Sarvam opens Token Factory to everyone: eight open-weight models - Sarvam 105B, GLM 5.3, DeepSeek V4.1 Flash, Kimi K3.1 Pro, Gemma 4 31B and more - served from GPU clusters in India through one OpenAI-compatible API, billed in rupees.
Why it matters
India-hosted inference with data residency inside the country, rupee billing and a single API across rival model families - Sarvam claims India's largest Nvidia B200 cluster behind it.
What you could build with it
An Indian health-tech startup could run patient-facing chat and document summarization on DeepSeek V4.1 Flash hosted inside India, keeping health data resident in-country while paying in rupees instead of dollars.
Does it hold up?
Open to everyone now with eight models behind one OpenAI-compatible API and rupee billing; Sarvam's latency and throughput figures are self-reported and untested by third parties, but the data-residency pitch is concrete for Indian teams.
Built with Sarvam Token Factory
- Sarvam Launches Sarvam Token Factory: Bringing India-Hosted Inference to 8 Open Modelsmarktechpost · Model catalogue and pricing: Sarvam 105B at 10 rupees/25 rupees per 1M tokens introductory, 200ms first-token latency, 330 tokens/sec claimed.
- Sarvam Token Factory Brings AI Hosting to Indiatechnosports · Launch explainer: OpenAI-compatible Chat Completions and Responses APIs, so teams migrate by swapping base URL and key.
- Sarvam AI opens India-hosted Token Factorytheoutpost · Competitive context: Krutrim has hosted DeepSeek in India since January 2025, and OpenAI began India inference for GPT-6 Luna on Nov 12.
Learn more
First spotted on article: source.
More AI infrastructure
vLLM v0.34.0: extensible KV cache by default, Model Runner V2 only, and 10 new modelsThe open inference engine's latest release spans 1,061 commits from 412 contributors, adds 10 models including Mistral…infra · JEV 0.86llama.cpp v1.0.0: the local-AI engine's first stable releasellama.cpp and ggml hit v1.0.0, promising semantic versioning so breaking C API changes only arrive with major bumps.infra · JEV 0.84colibri: run 744B-parameter MoE models on consumer hardwarePure-C inference engine with zero dependencies that streams MoE experts from NVMe disk, keeping hot experts on GPU…infra · JEV 0.78claude-mem: persistent memory plugin that lets coding agents pick up where the last session left offPlugin that hooks five Claude Code lifecycle events, compresses session activity into a local SQLite + Chroma store…infra · JEV 0.74FreeToken: datacenter-scale MoE serving on desktop GPUsAn edge-native mixture-of-experts serving engine from UC Berkeley researchers that runs 290B+ parameter MoE models on…infra · JEV 0.74GBrain: Garry Tan's self-wiring memory layer for AI agentsY Combinator CEO Garry Tan's open-source, markdown-first knowledge layer that gives OpenClaw and Hermes agents…infra · JEV 0.71
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.