S

Sarvam M

May 2025
SSarvam M
Compare in
on
—per task?
$0.70 / $1.16per 1M in / out?
—output speed?
8Kcontext

Ranking?

Strong at

C
Tool calling#229 of 242Reasoning#328 of 334Science#275 of 312Knowledge#265 of 318Instructions#204 of 245

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperMiMo v2.6 ProthinkingOpen×8.7-33%$0.13Memory (4-bit): 14 GB → 615 GB+4BetterGPT-6 Lunalow×4.1-75%$0.005Function calling only with none · Chat Completions+3OpenQwen3.8 27Bxhigh×6.3+17%$1.01Memory (4-bit): 14 GB → 17 GB+5CheaperDeepSeek V4.1 Flash 1017maxOpen×8.5-36%$0.26Memory (4-bit): 14 GB → 458 GB+5CheaperClaude Haiku 5.5max×8.2-75%$0.21Remove sampling parameters+5

Run it yourself

estimate
Fits on✗ 8 GB (phone, laptop)✓ 16 GB · compressed✓ 24 GB GPU · compressed✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
LicenceApache-2.0free for commercial use

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam3.1#331
AA Omniscience Index-61.6#266
CritPt (physics)0#180
AA Intelligence Index 4.3.2≈5.3#311
AA Omniscience · Accuracy15.2#229
AA Omniscience · Hallucination Rate90.6#254
GPQA Diamond41.6#278
Coding
Terminal-Bench Hard (version unstated)2.3#183
LiveCodeBench · AA archived29.5#122
Agents & Tool Use
τ²-bench · AA0#232
Writing & Language
IFBench (instructions)31.8#206

FAQ

How much does Sarvam M cost?›

$0.70 per 1M input tokens and $1.16 per 1M output tokens (cheapest host: Featherless).

Is Sarvam M free to use?›

Yes, there is a free API option.

How good is Sarvam M, and what is it best at?›

#328 of 334 in Chat & Knowledge, #181 of 239 in Coding, #229 of 242 in Agents & Tool Use, #204 of 245 in Writing & Language.

Is Sarvam M better than Ministral 3 8B 2512?›

On AA Intelligence Index 4.3.2, Sarvam M scores 5.3 vs 5.5 — about the same.

How can I use Sarvam M?›

Through the API of Featherless, Sarvam.

What is the context window of Sarvam M?›

8K tokens, with up to 32K output tokens.

When was Sarvam M released? Is there a newer version?›

Released 2025-05-23. It has been superseded by Sarvam 30B.

Is Sarvam M open source?›

Yes, the weights are open (Apache-2.0 licence).

Can I run Sarvam M locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

Can Sarvam M read images?›

No, it accepts text only.

What is Sarvam M?›

Sarvam M is an open-weights language model from Sarvam, released May 2025.

What are the best alternatives to Sarvam M?›

MiMo v2.6 Pro (cheaper), GPT-6 Luna (better), Qwen3.8 27B (open weights).