Compare / Nemotron 3 Nano 30B-A3B vs EXAONE 4.0 32B
Nemotron 3 Nano 30B-A3B vs EXAONE 4.0 32B
Updated November 2026
Verdict
Nemotron 3 Nano 30B-A3B and EXAONE 4.0 32B are roughly tied on the Small LM Index (62.6 vs 62.0). On the 4 benchmarks both labs report, Nemotron 3 Nano 30B-A3B wins 2 and EXAONE 4.0 32B wins 2. Both need about 21.9 GB of memory at 4-bit quantization. Only Nemotron 3 Nano 30B-A3B has a hosted API on OpenRouter ($0.10 per million tokens); the other has to be self-hosted.
Pick Nemotron 3 Nano 30B-A3B if…
- you mostly write code
- you need math and reasoning
- you want a hosted API rather than running it yourself
- you work with long documents
- you want the more widely used model (more quantized builds and community support)
Pick EXAONE 4.0 32B if…
No clear advantage in the data we have.
Specifications
| Nemotron 3 Nano 30B-A3B | EXAONE 4.0 32B | |
|---|---|---|
| Small LM Index | 62.6 (likely 59–66) | 62.0 (likely 59–65) |
| Lab | NVIDIA · United States | LG AI Research (EXAONE) · Rest of Asia |
| Released | 2025-12-15 | 2025-07-15 |
| Parameters | 31.6B (3.5B active, MoE) | 32B (dense) |
| Memory at 4-bit | ~20.3 GB | ~21.9 GB |
| Context window | 1M tokens | 131K tokens |
| Input | text | text |
| Reasoning | Hybrid (thinking on/off) | Hybrid (thinking on/off) |
| License | NVIDIA Open Model | EXAONE License (non-commercial) |
| API price / 1M tokens | $0.10 ($0.060 in, $0.24 out) | Self-host only |
| Hugging Face downloads (30 days) | 1.7M | 14.8K |
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 | Nemotron 3 Nano 30B-A3B | EXAONE 4.0 32B |
|---|---|---|
| MMLU-Pro Knowledge | 78.3 | 81.8 |
| GPQA Diamond Knowledge | 73.0 | 75.4 |
| Humanity's Last Exam Reasoning | 10.6 | ≈ 12 |
| AIME Math | 89.1 | 85.3 |
| LiveCodeBench Coding | 68.3 | 66.7 |
| SWE-bench Verified Coding | 38.8 | ≈ 49 |
| τ²-Bench Agentic | 49.0 | — |
| IFBench Instruction following | 71.5 | — |
| MMLU Knowledge | — | 92.3 |
| Multilingual MMLU Multilingual | — | 85.6 |
| IFEval Instruction following | — | 83.7 |
| BFCL Agentic | — | 63.9 |
| HMMT Math | — | 72.9 |
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, Nemotron 3 Nano 30B-A3B or EXAONE 4.0 32B?
Nemotron 3 Nano 30B-A3B and EXAONE 4.0 32B are roughly tied on the Small LM Index (62.6 vs 62.0). On the 4 benchmarks both labs report, Nemotron 3 Nano 30B-A3B wins 2 and EXAONE 4.0 32B wins 2.
Which is cheaper to run?
Only Nemotron 3 Nano 30B-A3B has a hosted API on OpenRouter ($0.10 per million tokens); the other has to be self-hosted. Both need about 21.9 GB of memory at 4-bit quantization.
Which has the longer context window?
Nemotron 3 Nano 30B-A3B has the longer context window: 1M tokens against 131K.
Can I use Nemotron 3 Nano 30B-A3B and EXAONE 4.0 32B commercially?
Nemotron 3 Nano 30B-A3B is released under NVIDIA Open Model and EXAONE 4.0 32B under EXAONE License (non-commercial). Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.