---
title: "Sarvam serves 450bn+ DeepSeek tokens a day in India, 95.7% of prompt tokens from cache"
summary: "Sarvam’s launch post and price pages of 16 November: open-weight models served from GPUs in India, what its cache does to the bill, and the rupee prices."
url: https://www.thetechbag.com/blog/sarvam-token-factory-450bn-tokens-deepseek
format: Note
desk: "AI in the enterprise"
region: India
published: 2026-11-17
facts_checked: 2026-11-17
author: TechBag (The TechBag Journal)
tags: ["Sarvam Token Factory","Sarvam Token Factory price","DeepSeek API India","DeepSeek V4.1 Flash price","Sarvam AI","Sarvam","DeepSeek V4.1 Flash","GLM 5.3","Kimi K3.1 Pro","Gemma 4 31B","Sarvam 105B","prompt caching","cached input tokens","AI inference India","open-weight models","B200","per million tokens","Mooncake"]
cover: https://www.thetechbag.com/blog/covers/sarvam-token-factory-450bn-tokens-deepseek.png
---

# Sarvam serves 450bn+ DeepSeek tokens a day in India, 95.7% of prompt tokens from cache

*Sarvam’s launch post and price pages of 16 November: open-weight models served from GPUs in India, what its cache does to the bill, and the rupee prices.*

Note · AI in the enterprise · By TechBag · Published 2026-11-17 · Facts checked 2026-11-17 · Canonical: https://www.thetechbag.com/blog/sarvam-token-factory-450bn-tokens-deepseek

Sarvam serves more than 450bn tokens a day of DeepSeek V4.1 Flash, an open-weight model from the Chinese lab DeepSeek, from GPU clusters in India, and 95.7% of the prompt tokens it handles for that model are served from cache rather than computed again. It gave both figures on 16 November, opening its inference service, Token Factory, to everyone.[1] It sells the model at ₹14 per million input tokens and ₹55 per million output tokens on a Standard endpoint, and ₹25 and ₹105 on an Interactive one built to hold about 200 tokens a second per user. Input that is reused from a cached prompt costs ₹0.25 per million on Standard and ₹0.50 on Interactive. All prices exclude GST.[2][3]

450bn+

tokens a day of DeepSeek V4.1 Flash, served from GPUs in India

Sarvam

95.7%

of that model’s prompt tokens served from cache

Sarvam

₹14 / ₹55

per million input / output tokens, Standard endpoint, introductory, excluding GST

Sarvam

₹0.25

per million cached input tokens, a fifty-sixth of the fresh rate

Sarvam

## What Token Factory sells

Token Factory serves Sarvam’s own Sarvam 105B alongside other labs’ open-weight models through one OpenAI-compatible API: GLM 5.3 and GLM 5.3 Flash from Z.ai, DeepSeek V4.1 Flash and V4 Pro, Kimi K3.1 Pro from Moonshot AI and Google’s Gemma 4 31B.[1] Its catalogue also lists Alibaba’s Qwen 3.8 27B and Sarvam’s Prism decision model. Inference runs on Sarvam-operated GPUs in India, and requests “are not forwarded to the original model developer’s overseas API”.[4] Sarvam says it runs “India’s largest B200 cluster” (the B200 is an Nvidia data-centre GPU); it gives no count of the chips.[1] It calls DeepSeek V4.1 Flash “one of the most used models on OpenRouter”, the gateway where the Journal tracked [Chinese models’ rising share of tokens](https://www.thetechbag.com/blog/chinese-open-weight-models-gateway-traffic).[1]

## Why the cache sets the bill

Every request runs in two phases: prefill reads the whole prompt in one compute-heavy pass, then decode writes the answer a token at a time. An AI agent calls a tool and comes back seconds or minutes later with almost the same prompt; Sarvam’s agent prompts on GLM 5.3 average 131,000 tokens. GPU memory held that context for about four seconds, two under load, so nearly every returning turn was read again from scratch. Sarvam added a shared store, built on Moonshot’s Mooncake, that every prefill server can read and that keeps prompts for hours or days. Over a week of production traffic its hit rate on DeepSeek rose from 91.4% to 95.7%, which halved the tokens that needed prefill; on GLM 5.3 at peak it cut the prefill servers needed from 58 to 34.[1]

**Figure: Cached input costs ₹0.25 per million tokens on Sarvam’s DeepSeek endpoint, against ₹14 fresh**

*Sarvam Token Factory, DeepSeek V4.1 Flash 1017: price per million tokens by endpoint and token type, ₹, excluding GST*

Standard: cached input

₹0.25

Standard: input

₹14

Standard: output

₹55

Interactive: cached input

₹0.5

Interactive: input

₹25

Interactive: output

₹105

|  | Value (₹) |
| --- | --- |
| Standard: cached input | 0.25 |
| Standard: input | 14 |
| Standard: output | 55 |
| Interactive: cached input | 0.5 |
| Interactive: input | 25 |
| Interactive: output | 105 |

Source: Sarvam Token Factory pricing page and DeepSeek V4.1 Flash 1017 model page · as of 17 November 2026 · Standard prices are introductory; reasoning tokens are billed as output

TechBag countWhat 95.7% does to an input bill

At Sarvam’s Standard rates, a million prompt tokens of which 95.7% are cached cost about ₹0.84 before GST, against ₹14 if none were: 957,000 tokens at ₹0.25 per million plus 43,000 at ₹14. A customer gets that rate only where its own prompts repeat a cached prefix, such as a long fixed system prompt or an agent’s loop; Sarvam’s 95.7% is the average across its traffic.[2]

Sarvam’s own comparison puts the Standard endpoint’s blended price at US$0.092 per million tokens, the same as DeepSeek’s own API, at 130 output tokens a second against 60, measured by Sarvam from Mumbai.[1] The one OpenAI model its API processes in India, [GPT-6 Luna](https://www.thetechbag.com/blog/openai-api-inference-india-residency-10-percent), lists at US$0.11 in and US$0.55 out per million with its 10% regional premium, and prices cached input at a tenth of fresh input; Sarvam prices it at a fifty-sixth.[5] Sarvam labels the Standard prices introductory and has not said when they end.[3]

How we did this

Figures as of 17 November 2026, 10:30pm IST

Volumes, cache rates, speeds and the blended prices are Sarvam’s own figures from its launch post; the blended-price formula is not stated. Its cache chart counts 872bn DeepSeek prompt tokens over 30 October to 6 November, a different measure and period from the 450bn-a-day figure, which Sarvam does not break down. Rupee prices are from the pricing and model pages as read on 17 November, excluding GST (18% is added on Indian invoices). Our arithmetic: the ₹0.84 example; “a fifty-sixth” (₹0.25 against ₹14); the comparison with GPT-6 Luna, whose list prices (US$0.10 in, US$0.01 cached, US$0.50 out per million) and 10% premium are from OpenAI’s model page. At Sarvam’s own conversion of ₹94 to the dollar, ₹14 and ₹55 are about US$0.15 and US$0.59. The models differ, so this compares prices only, not quality.

Spotted an error? Write to us; corrections are published at the foot of the piece.

Sources

1. [Sarvam, Introducing Sarvam Token Factory](https://www.sarvam.ai/blogs/sarvam-token-factory) — 16 November 2026
2. [Sarvam Token Factory, Pricing (INR per million tokens, excluding GST)](https://www.sarvam.ai/token-factory/pricing) — undated page, as read 17 November 2026
3. [Sarvam Token Factory, DeepSeek V4.1 Flash 1017 model page (Standard and Interactive endpoints)](https://www.sarvam.ai/token-factory/models/deepseek-v4-1-flash-1017) — undated page, as read 17 November 2026
4. [Sarvam Token Factory, Models (catalogue; where inference is processed)](https://www.sarvam.ai/token-factory/models) — undated page, as read 17 November 2026
5. [OpenAI, GPT-6 Luna model page (pricing; regional processing premium)](https://developers.openai.com/api/docs/models/gpt-6-luna) — undated page, as read 17 November 2026

**Disclosure.** TechBag is a software reseller: it earns when companies buy through it, including from vendors named here. The analysis is TechBag’s own; no vendor paid for or saw this piece before publication. Facts are labelled reported, TechBag count or our view.

**Read this next:** [US$300bn pledged to India's data centres; subsidised GPUs reportedly 20% used](https://www.thetechbag.com/blog/india-ai-compute-pledges-gpu-use) — The rest of India’s GPU build-out: US$300bn of data-centre pledges, and GPUs onboarded under the IndiaAI Mission reportedly about 20% used.

Go deeper on TechBag

- [OpenAI’s API now runs inside India: one model, at a 10% premium](https://www.thetechbag.com/blog/openai-api-inference-india-residency-10-percent) — the other way to keep model inference in India, and its price
- [Sarvam’s Prism: input at ₹4 per million tokens](https://www.thetechbag.com/blog/sarvam-prism-decision-model-4-rupees-million-tokens) — Sarvam’s own decision model, now in the same catalogue
- [Chinese AI models rose from 6–13% to 57–67% of OpenRouter tokens](https://www.thetechbag.com/blog/chinese-open-weight-models-gateway-traffic) — how much traffic open-weight models such as DeepSeek’s now carry
