Compare / Sarvam 30B vs Qwen3 32B

Sarvam 30B vs Qwen3 32B

Updated November 2026

Verdict

Sarvam 30B scores higher on the Small LM Index: 61.6 vs 52.4. On the 1 benchmark both labs report, Sarvam 30B wins 1 and Qwen3 32B wins 0. Sarvam 30B is smaller: it needs about 20.8 GB of memory at 4-bit, against 23.2 GB for Qwen3 32B. Only Qwen3 32B has a hosted API on OpenRouter ($0.13 per million tokens); the other has to be self-hosted.

Pick Sarvam 30B if…

you want the strongest model overall
you need math and reasoning
you have limited memory (about 20.8 GB at 4-bit)

Pick Qwen3 32B if…

you want a hosted API rather than running it yourself
you want the more widely used model (more quantized builds and community support)

Specifications

	Sarvam 30B	Qwen3 32B
Small LM Index	61.6 (likely 58–65)	52.4 (likely 47–57)
Lab	Sarvam AI · Rest of Asia	Alibaba (Qwen) · China
Released	2026-03-03	2025-04-29
Parameters	32.2B (2.4B active, MoE)	32.8B (dense)
Memory at 4-bit	~20.8 GB	~23.2 GB
Context window	131K tokens	131K tokens
Input	text	text
Reasoning	Always thinks	Hybrid (thinking on/off)
License	Apache-2.0	Apache-2.0
API price / 1M tokens	Self-host only	$0.13 ($0.080 in, $0.28 out)
Hugging Face downloads (30 days)	5K	3M

Benchmarks

Higher is better. Bold marks the winner where both labs report a score. ≈ marks the index model's estimate for a score that wasn't reported, ✓ an independent measurement and * a score measured by another lab.

Benchmark	Sarvam 30B	Qwen3 32B
MMLU-Pro Knowledge	80.0	≈ 74
GPQA Diamond Knowledge	66.5	≈ 66
AIME Math	88.3	72.9
LiveCodeBench Coding	≈ 71	65.7
τ²-Bench Agentic	45.7	—
MMLU Knowledge	85.1	—
MATH-500 Math	97.0	—
HMMT Math	73.3	—

Scores are mostly reported by each lab with its own settings, so small gaps are noise. See the methodology. Open in the interactive comparison to add more models.

Questions

Which is better, Sarvam 30B or Qwen3 32B?

Sarvam 30B scores higher on the Small LM Index: 61.6 vs 52.4. On the 1 benchmark both labs report, Sarvam 30B wins 1 and Qwen3 32B wins 0.

Which is cheaper to run?

Only Qwen3 32B has a hosted API on OpenRouter ($0.13 per million tokens); the other has to be self-hosted. Sarvam 30B is smaller: it needs about 20.8 GB of memory at 4-bit, against 23.2 GB for Qwen3 32B.

Which has the longer context window?

Both have a 131K-token context window.

Can I use Sarvam 30B and Qwen3 32B commercially?

Both are released under the Apache-2.0 license. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.