whatmodel.sh / Sarvam-M

Sarvam-M

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Sarvam-M is an open-weights language model from Sarvam AI, released May 2025. Its published size is 23.57B parameters. Licence: apache-2.0.

Catalogue identity notes

Maker repo sarvamai/sarvam-m @01534a5: licence tag apache-2.0. 23,572,403,200 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2025-05: month the maker's Hugging Face repo was created (createdAt 2025-05-20).

37.7±5.0#760 of 949 LLMsfrom 1 benchmark result across 1 source* thinly tested, treat the rank as provisional

$ how it compares

The LLMs ranked either side of Sarvam-M, out of 1132 with a score.

#modelscorewhereRAMreleased
923Qwen3.5 0.8B Alibaba38±1local~2GB2026-03
924MAP-Neo 7B Instruct v0.1 ExPO Chujie Zheng (ExPO)38±3local~18GB2024-05
925Sarvam-M Sarvam AI38±5local~16GB2025-05
926Code Tulu 2 7B Allen Institute for AI38±2local~16GB2023-11
927zephyr-7b-alpha-ExPO Chujie Zheng (ExPO)38±2local~6.1GB2024-04

see all LLMs · find the ones that fit your machine

$ what it's good at

other tests1 not in a task · 89% of its score
testresultvs the others in its rundifficultycountswhy
HindiGen-v1 3C3H score (Inception, Claude 3.5 Sonnet judge)22.9% ✎▁▁▁▃·▂·█▆█▇▁ beats 3/31medium · avg 48%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 89%at its level

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run Sarvam-M · needs ~14.3 GB · from huggingface.co/lmstudio-community
    chat in the terminal
    llama cli -hf lmstudio-community/sarvam-m-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf lmstudio-community/sarvam-m-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Sarvam-M · needs ~14.3 GB · from huggingface.co/lmstudio-community

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/lmstudio-community/sarvam-m-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Sarvam-M · needs ~14.3 GB · from huggingface.co/lmstudio-community
    chat in the terminal
    ollama run hf.co/lmstudio-community/sarvam-m-GGUF:Q4_K_M

MLX (Mac)

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Sarvam-M · needs ~13.3 GB, Mac only · from huggingface.co/Jimmi42
    chat in the terminal
    mlx_lm.chat --model Jimmi42/sarvam-m-4bit-mlx
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model Jimmi42/sarvam-m-4bit-mlx

Other ways to run it

  1. Run it as published · needs ~47.14 GB · from huggingface.co/sarvamai (official)
    uv run --with torch --with 'transformers<5' --with accelerate python -c "from transformers import pipeline; print(pipeline('text-generation', model='sarvamai/sarvam-m', torch_dtype='auto', device_map='auto')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"

Official weights and documentation: huggingface.co/sarvamai ↗

$ how the score was worked out · Inception Arabic/Hindi leader… 89%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval)1▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 89%Apache-2.0
prior · pull toward the middle▇▇▇ 11%
weighted average= 37.7

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json