Compare / Spark-X2.6 1.7B vs EXAONE 4.0 1.2B
Spark-X2.6 1.7B vs EXAONE 4.0 1.2B
Updated November 2026
Verdict
Spark-X2.6 1.7B scores higher on the Small LM Index: 50.1 vs 36.1. On the 5 benchmarks both labs report, Spark-X2.6 1.7B wins 5 and EXAONE 4.0 1.2B wins 0. Both need about 1.8 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick Spark-X2.6 1.7B if…
- you want the strongest model overall
- you need math and reasoning
- you need precise instruction following
- you build agents that call tools
- you work with long documents
- you need a permissive license (Apache-2.0)
Pick EXAONE 4.0 1.2B if…
- you want the more widely used model (more quantized builds and community support)
Specifications
| Spark-X2.6 1.7B | EXAONE 4.0 1.2B | |
|---|---|---|
| Small LM Index | 50.1 (likely 47–53) | 36.1 (likely 33–40) |
| Lab | SparkLLM (XHToken) · China | LG AI Research (EXAONE) · Rest of Asia |
| Released | 2026-10-19 | 2025-07-15 |
| Parameters | 1.7B (dense) | 1.3B (dense) |
| Memory at 4-bit | ~1.7 GB | ~1.8 GB |
| Context window | 1M tokens | 64K tokens |
| Input | text | text |
| Reasoning | Hybrid (thinking on/off) | Hybrid (thinking on/off) |
| License | Apache-2.0 | EXAONE License (non-commercial) |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 2.7K | 13.3K |
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 | Spark-X2.6 1.7B | EXAONE 4.0 1.2B |
|---|---|---|
| MMLU-Pro Knowledge | ≈ 72 | 59.3 |
| GPQA Diamond Knowledge | 56.1 | 52.0 |
| Humanity's Last Exam Reasoning | 8.7 | ≈ 3 |
| AIME Math | 70.1 | 45.2 |
| LiveCodeBench Coding | ≈ 57 | 45.3 |
| SWE-bench Verified Coding | 59.8 | ≈ 17 |
| τ²-Bench Agentic | 85.3 | — |
| IFBench Instruction following | 65.5 | — |
| MMLU Knowledge | — | 71.5 |
| Multilingual MMLU Multilingual | — | 62.4 |
| IFEval Instruction following | 89.4 | 67.8 |
| BFCL Agentic | 55.2 | 52.9 |
| HMMT Math | 48.8 | 34.0 |
| AA-LCR Long context | 48.0 | — |
| SciCode Coding | 20.1 | — |
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, Spark-X2.6 1.7B or EXAONE 4.0 1.2B?
Spark-X2.6 1.7B scores higher on the Small LM Index: 50.1 vs 36.1. On the 5 benchmarks both labs report, Spark-X2.6 1.7B wins 5 and EXAONE 4.0 1.2B wins 0.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 1.8 GB of memory at 4-bit quantization.
Which has the longer context window?
Spark-X2.6 1.7B has the longer context window: 1M tokens against 64K.
Can I use Spark-X2.6 1.7B and EXAONE 4.0 1.2B commercially?
Spark-X2.6 1.7B is released under Apache-2.0 and EXAONE 4.0 1.2B 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.