S

Sarvam 105B

Mar 2026
SSarvam 105B
Compare in
on
—per task?
$15 / $60per 1M in / out
212 tok/sfirst token in 1.6 s?
128Kcontext

Ranking?

Strong at

B
Tool calling#115 of 242
C
Reasoning#173 of 334Science#160 of 312Long context#269 of 273Knowledge#255 of 318Instructions#187 of 245

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperClaude Opus 5.5max×6.6-70%$5.98No forced tool use+5BetterGPT-6 Lunalow×2.4-99%$0.005Function calling only with none · Chat Completions+3OpenQwen3.8 Flashthinking×4.5-99%$0.37Memory (4-bit): 64 GB → 108 GB+6CheaperClaude Sonnet 5.5max×6.4-85%$5.46No forced tool use+5CheaperClaude Fable 5.1max×6.1-24%$7.63Replace forced tool_choice+5

Run it yourself

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

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam11#174
AA Omniscience Index-59.4#256
CritPt (physics)0#180
AA Intelligence Index 4.3.2≈8.8#236
AA Omniscience · Accuracy17.6#192
AA Omniscience · Hallucination Rate93.4#292
AA-LCR (long context) 1.10#271
GPQA Diamond73.8#162
Coding
Terminal-Bench Hard (version unstated)1.5#197
Agents & Tool Use
τ²-bench · AA46.8#117
Writing & Language
IFBench (instructions)34.4#189

FAQ

How much does Sarvam 105B cost?›

$15 per 1M input tokens and $60 per 1M output tokens.

Is Sarvam 105B free to use?›

There is no free API; the cheapest paid price is $15 per 1M input tokens. The weights are open, so you can run it yourself.

How good is Sarvam 105B, and what is it best at?›

#173 of 334 in Chat & Knowledge, #195 of 239 in Coding, #115 of 242 in Agents & Tool Use, #187 of 245 in Writing & Language.

Is Sarvam 105B better than MiniCPM5 1B?›

On AA Intelligence Index 4.3.2, Sarvam 105B scores 8.8 vs 8.8 — about the same.

How can I use Sarvam 105B?›

Through the Sarvam API.

What is the context window of Sarvam 105B?›

128K tokens.

When was Sarvam 105B released? Is there a newer version?›

Released 2026-03-06. It is the current version in its line.

Is Sarvam 105B open source?›

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

Can I run Sarvam 105B locally?›

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

Can Sarvam 105B read images?›

No, it accepts text only.

What is Sarvam 105B?›

Sarvam 105B is an open-weights language model from Sarvam, released Mar 2026.

What are the best alternatives to Sarvam 105B?›

Claude Opus 5.5 (cheaper), GPT-6 Luna (better), Qwen3.8 Flash (open weights).