NOUMENOpen app
NOUMEN

sarvam · Efficient

Sarvam 105B

Added May 12, 2026

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Sarvam 105B is a 105B parameter chat completion model from Sarvam AI with multilingual support, streaming, tool calling, reasoning controls, and a 128k context window.

  • Reasoning
  • Tools
Input
Text
Output
Text
Reasoning effort
low · medium · high
Discovery tier
Efficient

Pricing

USD per 1M tokens
Input
$0.054
Output
$0.21
Cache read
$0.0336

Reference prices as published by the model provider. API example

Specifications

Context window
131.1K
Max output
4.1K
Average output (7 days)
1.1K tokens
Longer than 74% of 468 models

Side by side

Sarvam 105B compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Sarvam 105BsarvamThis modelGemini 2.0 Pro 0205GoogleRemove
ProvidersarvamGoogle
Discovery tierEfficientBalanced
AddedMay 12, 2026Feb 5, 2025
Limits
Context window131.07K2.1MBest
Max output4.1K8.19KBest
Average output1.1K tokens—
Pricing, USD per 1M tokens
Input$0.054Best$1.99
Output$0.21Best$7.96
Cache read$0.0336Best$0.50
Benchmark indexes
Intelligence Index8.8Best8.7
Coding Index—25.5
Capabilities
ReasoningYesNo
ToolsYesNo
VisionNoYes
Audio inputNoYes
Reasoning effortlow, medium, high—

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Sarvam 105B
  • Gemini 2.0 Pro 0205

Intelligence Index

  • Sarvam 105B8.8
  • Gemini 2.0 Pro 02058.7

Sarvam 105B tested at high reasoning effort

Sarvam 105B: Better than 16% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Sarvam 105B: —
    • Gemini 2.0 Pro 0205: 25.5

Document reasoning

  • AA-LCRv1.1

    Reasoning across long documents

    • Sarvam 105B: 0.0%
    • Gemini 2.0 Pro 0205: —

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Sarvam 105B: 11.0%Top
    • Gemini 2.0 Pro 0205: 6.2%

  • GPQA Diamond

    Graduate-level science questions

    • Sarvam 105B: 73.8%Top
    • Gemini 2.0 Pro 0205: 62.2%

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Sarvam 105B: —
    • Gemini 2.0 Pro 0205: 80.5%

  • IFBench

    Following precise output instructions

    • Sarvam 105B: 34.4%
    • Gemini 2.0 Pro 0205: —

Coding

  • LiveCodeBench

    Recent competitive programming problems

    • Sarvam 105B: —
    • Gemini 2.0 Pro 0205: 34.7%

Math

  • AIME

    Competition mathematics

    • Sarvam 105B: —
    • Gemini 2.0 Pro 0205: 36.0%

  • MATH-500

    Competition math problems

    • Sarvam 105B: —
    • Gemini 2.0 Pro 0205: 92.3%

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Sarvam 105B: 1.5%
    • Gemini 2.0 Pro 0205: —

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Sarvam 105B: 46.8%
    • Gemini 2.0 Pro 0205: —

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Sarvam 105B: 0.0%
    • Gemini 2.0 Pro 0205: —

Scores from Artificial Analysis, updated Oct 26, 2026. Sarvam 105B was tested at high reasoning effort; compare scores measured at the same effort.Artificial Analysis

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sarvam-105b",
    "reasoning_effort": "medium",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="sarvam-105b",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)