Compare / Spark-X2.6 4B vs Spark-X2.5 4B
Spark-X2.6 4B vs Spark-X2.5 4B
Updated November 2026
Verdict
Spark-X2.6 4B scores higher on the Small LM Index: 67.5 vs 65.5, although their likely ranges overlap, so the lead is not certain. On the 11 benchmarks both labs report, Spark-X2.6 4B wins 11 and Spark-X2.5 4B wins 0. Both need about 3.4 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 4B if…
- you want the strongest model overall
- you mostly write code
- you need math and reasoning
- you need precise instruction following
- you build agents that call tools
Pick Spark-X2.5 4B if…
No clear advantage in the data we have.
Specifications
| Spark-X2.6 4B | Spark-X2.5 4B | |
|---|---|---|
| Small LM Index | 67.5 (likely 65–70) | 65.5 (likely 63–68) |
| Lab | SparkLLM (XHToken) · China | SparkLLM (XHToken) · China |
| Released | 2026-10-19 | 2026-08-24 |
| Parameters | 4.1B (dense) | 4.1B (dense) |
| Memory at 4-bit | ~3.4 GB | ~3.4 GB |
| Context window | 1M tokens | 1M tokens |
| Input | text | text |
| Reasoning | Hybrid (thinking on/off) | Hybrid (thinking on/off) |
| License | Apache-2.0 | Apache-2.0 |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 18.9K | 30.3K |
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 4B | Spark-X2.5 4B |
|---|---|---|
| GPQA Diamond Knowledge | 77.1 | 73.7 |
| Humanity's Last Exam Reasoning | 17.9 | 12.3 |
| AIME Math | 91.2 | 90.7 |
| SWE-bench Verified Coding | 62.8 | 41.6 |
| τ²-Bench Agentic | 86.9 | 75.1 |
| IFBench Instruction following | 75.8 | 75.0 |
| IFEval Instruction following | 93.6 | 93.0 |
| BFCL Agentic | 71.8 | 65.1 |
| HMMT Math | 81.3 | 81.2 |
| AA-LCR Long context | 67.3 | 56.3 |
| SciCode Coding | 36.8 | 34.7 |
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 4B or Spark-X2.5 4B?
Spark-X2.6 4B scores higher on the Small LM Index: 67.5 vs 65.5, although their likely ranges overlap, so the lead is not certain. On the 11 benchmarks both labs report, Spark-X2.6 4B wins 11 and Spark-X2.5 4B wins 0.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 3.4 GB of memory at 4-bit quantization.
Which has the longer context window?
Both have a 1M-token context window.
Can I use Spark-X2.6 4B and Spark-X2.5 4B 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.
Related comparisons
- Nanbeige4.2 3B vs Spark-X2.6 4B
- Qwen3.5 4B vs Spark-X2.6 4B
- Qwen3 4B Thinking 2507 vs Spark-X2.6 4B
- Nemotron Labs Dory 1.0 4B vs Spark-X2.6 4B
- Gemma 4 E2B vs Spark-X2.6 4B
- Gemma 3 4B vs Spark-X2.6 4B
- Gemma 3n E2B vs Spark-X2.6 4B
- Nanbeige4.2 3B vs Spark-X2.5 4B
- Qwen3.5 4B vs Spark-X2.5 4B
- Qwen3 4B Thinking 2507 vs Spark-X2.5 4B