Model benchmarks
Spark-X2.6-4B-Q4_K_M local LLM performance
As of October 2026, Spark-X2.6-4B-Q4_K_M runs at up to 72.1 tok/s for local inference (best of 4 community benchmark runs across 1 GPU).
Model size
4B
Peak speed
72.1 tok/s
Average speed
69.0 tok/s
Avg PP
1526.2 tok/s
Min memory
4.8 GB
Max context
32,768 tokens
Avg output / run
34,416 tokens
Avg runtime / run
8m 36s
Avg quality
61.1
Benchmark runs
4
GPUs tested
1
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 4 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 54.8 | 61.1 | 67.1 |
| Agent Workflow | 59.3 | 78.4 | 90.3 |
| Code Generation | 10.7 | 34.6 | 50.3 |
| Role Play & Narrative | 62.6 | 65.1 | 67.6 |
| Research & Analysis | 40.1 | 66.1 | 87.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination Spark-X2.6-4B-Q4_K_M has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3060 | llama.cpp | Q4_K_M | 72.1 tok/s | 69.0 tok/s | 4.8 GB | 32,768 tokens | 61.1 | 4 |
Benchmark runs
All 4 Spark-X2.6-4B-Q4_K_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Spark-X2.6-4B-Q4_K_M good for coding?
- In our benchmarks, Spark-X2.6-4B-Q4_K_M scores 34.6/100 for coding. It runs at about 69.0 tok/s, so if you want more speed, Nex-N2.5-mini-IQ3_XXS is faster (~91.4 tok/s) and still scores well for coding (81.6/100).
- Is Spark-X2.6-4B-Q4_K_M good for agentic (tool-using) tasks?
- In our benchmarks, Spark-X2.6-4B-Q4_K_M scores 78.4/100 for agentic workflows. It runs at about 69.0 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
- How fast is Spark-X2.6-4B-Q4_K_M for local inference?
- Across 4 community benchmark runs, Spark-X2.6-4B-Q4_K_M reaches up to 72.1 tok/s and averages 69.0 tok/s, with the fastest results on NVIDIA GeForce RTX 3060.
- How much memory does Spark-X2.6-4B-Q4_K_M need?
- The leanest observed configuration used about 4.8 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run Spark-X2.6-4B-Q4_K_M?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.