Sarvam AI · FinOps Profile

Sarvam Ai Finops

FinOps view of Sarvam AI spend. Sarvam bills usage-based per-token rates for chat LLM (separate input, cached-input, and output meters per model), per-hour rates for speech-to-text and speech-to-text-translate, and per-10,000-character rates for text-to-speech, translation, transliteration, and language identification. Published rates are in Indian Rupees (INR); accounts start with free credits.

Sarvam Ai Finops is the FinOps profile for Sarvam AI on the APIs.io network, aligned with the FinOps Foundation Framework.

It defines 7 billable meters, billed in INR, on a monthly cycle, and pricing category usage-based.

The profile maps 8 FOCUS columns for cost-allocation reporting.

Tagged areas include AI, LLM, Speech to Text, Text to Speech, and Translation.

Category: AI and Machine Learning Pricing: Usage-Based Billing: Monthly FOCUS v1.3
AILLMSpeech to TextText to SpeechTranslationIndian LanguagesFinOpsCost ManagementFOCUS

Framework Alignment

Framework
FinOps Foundation Framework
Data Spec
FOCUS v1.3

Charge Categories

UsagePurchaseAdjustment

FOCUS Columns

BillingCurrency
INR
ChargeCategory
Usage
InvoiceIssuerName
Sarvam AI
PricingCategory
Usage-Based
ProviderName
Sarvam AI
PublisherName
Sarvam AI
ServiceCategory
AI and Machine Learning
ServiceName
Sarvam AI API

Meters

input_tokens
Unit: tokens
Tokens sent in chat completion requests, billed per 1M tokens per model.
cached_input_tokens
Unit: tokens
Cached-input tokens billed at the discounted cached-input rate per model.
output_tokens
Unit: tokens
Tokens generated, billed per 1M tokens per model.
stt_audio_hours
Unit: hours
Audio hours transcribed via speech-to-text and speech-to-text-translate, billed per hour.
tts_characters
Unit: characters
Text-to-speech characters synthesized, billed per 10,000 characters per Bulbul model.
translate_characters
Unit: characters
Translation and transliteration characters, billed per 10,000 characters.
language_id_characters
Unit: characters
Language identification characters, billed per 10,000 characters.

Sources