Compare / EXAONE 4.0 32B vs Sarvam 30B

EXAONE 4.0 32B vs Sarvam 30B

Updated November 2026

Verdict

EXAONE 4.0 32B and Sarvam 30B are roughly tied on the Small LM Index (62.0 vs 61.6). On the 5 benchmarks both labs report, EXAONE 4.0 32B wins 3 and Sarvam 30B wins 2. Both need about 21.9 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.

Pick EXAONE 4.0 32B if…

No clear advantage in the data we have.

Pick Sarvam 30B if…

you need math and reasoning
you need a permissive license (Apache-2.0)

Specifications

	EXAONE 4.0 32B	Sarvam 30B
Small LM Index	62.0 (likely 59–65)	61.6 (likely 58–65)
Lab	LG AI Research (EXAONE) · Rest of Asia	Sarvam AI · Rest of Asia
Released	2025-07-15	2026-03-03
Parameters	32B (dense)	32.2B (2.4B active, MoE)
Memory at 4-bit	~21.9 GB	~20.8 GB
Context window	131K tokens	131K tokens
Input	text	text
Reasoning	Hybrid (thinking on/off)	Always thinks
License	EXAONE License (non-commercial)	Apache-2.0
API price / 1M tokens	Self-host only	Self-host only
Hugging Face downloads (30 days)	14.8K	5K

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	EXAONE 4.0 32B	Sarvam 30B
MMLU-Pro Knowledge	81.8	80.0
GPQA Diamond Knowledge	75.4	66.5
AIME Math	85.3	88.3
LiveCodeBench Coding	66.7	≈ 71
τ²-Bench Agentic	—	45.7
MMLU Knowledge	92.3	85.1
Multilingual MMLU Multilingual	85.6	—
IFEval Instruction following	83.7	—
BFCL Agentic	63.9	—
MATH-500 Math	—	97.0
HMMT Math	72.9	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, EXAONE 4.0 32B or Sarvam 30B?

EXAONE 4.0 32B and Sarvam 30B are roughly tied on the Small LM Index (62.0 vs 61.6). On the 5 benchmarks both labs report, EXAONE 4.0 32B wins 3 and Sarvam 30B wins 2.

Which is cheaper to run?

Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 21.9 GB of memory at 4-bit quantization.

Which has the longer context window?

Both have a 131K-token context window.

Can I use EXAONE 4.0 32B and Sarvam 30B commercially?

EXAONE 4.0 32B is released under EXAONE License (non-commercial) and Sarvam 30B under Apache-2.0. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.