How much does Sarvam: Sarvam 105B cost?

Sarvam: Sarvam 105B costs $0.04 per 1M input tokens and $0.16 per 1M output tokens. Prompt caching is priced with cache reads at $0.02/1M tokens.

What is the context length of Sarvam: Sarvam 105B?

Sarvam: Sarvam 105B has a 128,000 token context window.

Does Sarvam: Sarvam 105B support tool calling and structured outputs?

Yes. Sarvam: Sarvam 105B supports tool calling.

What inputs and outputs does Sarvam: Sarvam 105B support?

Sarvam: Sarvam 105B accepts text as input and returns text.

What other models does Sarvam have?

Sarvam: Bulbul V2, Sarvam: Saaras V3 and Sarvam: Sarvam 105B (Free) are other models from Sarvam.

When was Sarvam: Sarvam 105B released?

Sarvam: Sarvam 105B was released on Feb 18, 2026.

Sarvam 105B

SarvamReleased Feb 18, 2026sarvam/sarvam-105b

Compare

Sarvam 105B is Sarvam AI's flagship open-source Mixture-of-Experts (MoE) large language model (LLM), trained from scratch as India's first competitive 105-billion-parameter model. It delivers state-of-the-art performance across Indian languages and excels in enterprise-grade applications, with strong capabilities in multi-step reasoning, mathematics, coding, knowledge retrieval, and instruction-following.

Context

128K

Max output

128K

Input /1M

$0.04

Output /1M

$0.16

Blended /1M

$0.07

AcceptsTextProducesText

Providers

FastRouter routes your requests to this provider. You can also bring your own key.

ProviderInput /1MOutput /1MContextMax outputLatencyThroughput

Sarvam

sarvam

Input /1M$0.04

Output /1M$0.16

Context128K

Max output128K

Latency—

Throughput—

Supported parameters

Request parameters you can send with this model.

Core (Sampling & basic generation)

max_tokens: Maximum tokens to generate.
n: Number of completions to generate.
seed: Seed for deterministic sampling.
stop: Sequences where generation stops.
temperature: Controls randomness of output.
top_p: Nucleus sampling threshold.

Penalties (Token Penalties)

frequency_penalty: Penalizes frequent tokens.
presence_penalty: Penalizes tokens already present.

Reasoning (Reasoning Controls)

include_reasoning: Include reasoning trace in the response.
reasoning: Provider-specific reasoning configuration.
reasoning_effort: Reasoning effort before answering.

Tool Use (Function Calls & Tools)

tool_choice: How the model should use tools.
tools: Tool/function definitions available to the model.

Performance

Median throughput and time to first token per provider over the last week.

Throughput

Latency

Make your first API call

OpenAI-compatible. Point your SDK at FastRouter, or call the API directly.

Full API docs

Replace <FASTROUTER_API_KEY> with your key · Get your key →Install

OpenAI SDKHTTP

from openai import OpenAI

client = OpenAI(

base_url="https://api.fastrouter.ai/api/v1", # FastRouter base URL

api_key= "<FASTROUTER_API_KEY>", # Replace with your FastRouter API key

)

completion = client.chat.completions.create(

model="sarvam/sarvam-105b", # Replace with your model ID

messages=[

{ "role": "user", "content": "What is the meaning of life?" }

]

)

print(completion.choices[0].message.content)

Frequently asked questions

More models from Sarvam

Bulbul V2sarvam/bulbul:v2

Sarvam: Bulbul V2 is a text-to-speech (TTS) model from Sarvam AI, delivering real-time, natural-sounding speech in 11 Indian languages at a fraction of global costs. It builds on prior versions with key enhancements including improved audio quality, 30+ speaker voices (expanded from six distinct voices in v1 for diverse use cases like professional or conversational tones), and adjustable speech speed from 0.5x to 2.0x for customized delivery.

Saaras V3sarvam/saaras:v3

Saaras V3 is Sarvam AI's next-generation automatic speech recognition (ASR) model engineered for Indian languages and real-world speech conditions. Saaras V3 supports 23 languages (22 official Indian languages plus English) within a unified multilingual model. The model achieves approximately 19% Word Error Rate (WER) on the IndicVoices benchmark, improving significantly from its predecessor Saaras V2.5 which had ~22% WER. On the 10 most popular languages in the IndicVoices dataset, it achieves 19.3% WER, and performance advantages widen for lower-resource Indian languages

Sarvam 105B (Free)sarvam/sarvam-105b:free

Sarvam 105B is Sarvam AI's flagship open-source Mixture-of-Experts (MoE) large language model, trained from scratch as India's first competitive 105-billion-parameter model. It delivers strong performance across Indian languages and excels in enterprise-grade applications, with capabilities in multi-step reasoning, mathematics, coding, knowledge retrieval, and instruction-following.

Model scores sourced from ArtificialAnalysis