Skip to content

Head-to-head · Sarvam

Sarvam-30B vs Sarvam-105B

Sarvam-30B needs ~21.7 GB at Q4_K_M; Sarvam-105B needs ~67.5 GB. That ~45.8 GB gap decides which hardware runs each; they differ on 3 of the 14 devices below.

Spec Sarvam-30B Sarvam-105B
Parameters30B MoE105B MoE
Memory at Q4_K_M~21.7 GB~67.5 GB
Context window64k128k
LMArenanot rankednot ranked
LicenseApache-2.0Apache-2.0

Which devices run each

A representative spread across the memory range. Tap a verdict for the full breakdown.

Size at each quantization

Quant Sarvam-30B Sarvam-105B
Q2_K 12.6 GB* 44 GB*
Q3_K_M 14.7 GB* 51.3 GB*
Q4_K_M 19.6 GB 64.2 GB
Q5_K_M 21.4 GB* 74.8 GB*
Q6_K 24.6 GB* 86.1 GB*
Q8_0 31.9 GB* 111.6 GB*
FP16 60 GB* 210 GB*

* derived from bits-per-weight; unstarred sizes are measured GGUF files.

Bottom line

The larger Sarvam-105B needs ~67.5 GB; Sarvam-30B runs on lighter hardware at ~21.7 GB with more headroom and faster responses. Check each against your exact device: what runs Sarvam-30B · what runs Sarvam-105B.

More Sarvam comparisons

FAQ

What is the difference in memory between Sarvam-30B and Sarvam-105B?

At Q4_K_M, Sarvam-30B needs about 21.7 GB and Sarvam-105B needs about 67.5 GB, a difference of ~45.8 GB.

Should I run Sarvam-30B or Sarvam-105B?

Run Sarvam-105B if your hardware has the ~67.5 GB it needs and you want maximum quality; run Sarvam-30B (~21.7 GB) for lighter hardware, faster responses, and more memory headroom.

Full breakdowns: Sarvam-30B · Sarvam-105B · all models × devices.

Sources

Memory figures are estimates at Q4_K_M, validated 2026-08-03. See methodology.