# Sarvam AI — Introducing Sarvam Token Factory

- Company: Sarvam AI (sarvam.ai)
- Announced: 2026-11-07
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.sarvam.ai/blogs/sarvam-token-factory/
- Record: https://forck.live/items/36635-introducing-sarvam-token-factory
- Subject: Sarvam / Bulbul / Saaras
- Models affected: Sarvam 105B, Gemma 4 31B, GLM 5.3, DeepSeek V4.1 Flash, Kimi K3.1 Pro, Qwen 3.8 27B, Sarvam Prism, GLM 5.3 Flash, DeepSeek V4 Pro, Clef, EmbeddingGemma 2, Sarvam 105B Conversations, DeepSeek V4.1 Flash 1017, Kimi K3
- Pricing: DeepSeek V4.1 Flash introductory pricing ₹14 per million input tokens and ₹55 per million output tokens; Interactive ₹25 input and ₹105 output per million tokens, Standard ₹14 input and ₹55 output per million tokens

Sarvam is launching Sarvam Token Factory, an inference stack that serves its own models and leading open-weight models from GPU clusters in India through one OpenAI-compatible API. The company says that on most of the models it serves, it matches or beats the leading inference providers globally on both speed and cost. DeepSeek V4.1 Flash launches with introductory pricing of ₹14 per million input tokens and ₹55 per million output tokens, and is offered through an Interactive endpoint at about 200 tokens per second per user for coding agents and a Standard endpoint for background and batch work. Sarvam reports that Gemma 4 31B returns its first token in 0.40 seconds on a 10K-token prompt measured from Mumbai, 5× faster than the only other provider serving it, and streams at 308 tokens per second. On DeepSeek V4.1 Flash the company says it serves 450B+ tokens a day, with 95.7% of prompt tokens served from a three-tier cache. Sarvam says it runs India's largest B200 cluster and also runs the same inference optimization stack on-premise for critical institutions.

## Evidence

Verbatim from https://www.sarvam.ai/blogs/sarvam-token-factory/:

> Models served on Sarvam Token Factory include Sarvam 105B, Gemma 4 31B, GLM 5.3, DeepSeek V4.1 Flash, Kimi K3.1 Pro, Qwen 3.8 27B, Sarvam Prism, GLM 5.3 Flash, DeepSeek V4 Pro, Clef, EmbeddingGemma 2, Sarvam 105B Conversations .
> 
> Today, we are launching Sarvam Token Factory for everyone. Token Factory is our inference stack to serve our own models and leading open-weight models from GPU clusters located in India. On most of the models we serve, we match or beat the leading inference providers globally on both speed and cost.
> 
> Along with Sarvam 105B, we serve GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash 1017, DeepSeek V4 Pro, Kimi K3, Kimi K3.1 Pro and Gemma 4 31B through one OpenAI-compatible API. See the full model catalogue .
> 
> Sarvam Token Factory at a glance
> 
> Cost: We offer globally competitive pricing on every model we serve. DeepSeek V4.1 Flash launches with introductory pricing of ₹14 per million input tokens and ₹55 per million output tokens.
> 
> Performance:
> 
> Latency: Gemma 4 31B returns its first token in 0.40 seconds on a 10K-token prompt, 5× faster than the only other provider serving it.

---

Record: https://forck.live/items/36635-introducing-sarvam-token-factory
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md
