Model benchmarks
Jan-v3.5-4B-Q4_K_XL local LLM performance
As of October 2026, Jan-v3.5-4B-Q4_K_XL runs at up to 75.3 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
4B
Peak speed
75.3 tok/s
Average speed
73.4 tok/s
Avg PP
1965.7 tok/s
Min memory
7.1 GB
Max context
32,768 tokens
Avg output / run
8,434 tokens
Avg runtime / run
1m 57s
Avg quality
67.1
Benchmark runs
3
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 3 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 | 66.2 | 67.1 | 68.1 |
| Agent Workflow | 76.8 | 81.6 | 86.7 |
| Code Generation | 39.0 | 40.7 | 42.8 |
| Role Play & Narrative | 84.5 | 86.0 | 87.0 |
| Research & Analysis | 58.8 | 60.0 | 61.4 |
Performance by hardware and tool
Every hardware/tool/quantization combination Jan-v3.5-4B-Q4_K_XL 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 Laptop GPU | Jan | Q4_K | 75.3 tok/s | 73.4 tok/s | 7.1 GB | 32,768 tokens | 67.1 | 3 |
Benchmark runs
All 3 Jan-v3.5-4B-Q4_K_XL runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Jan-v3.5-4B-Q4_K_XL good for coding?
- In our benchmarks, Jan-v3.5-4B-Q4_K_XL scores 40.7/100 for coding. It runs at about 73.4 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 Jan-v3.5-4B-Q4_K_XL good for agentic (tool-using) tasks?
- In our benchmarks, Jan-v3.5-4B-Q4_K_XL scores 81.6/100 for agentic workflows. It runs at about 73.4 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 Jan-v3.5-4B-Q4_K_XL for local inference?
- Across 3 community benchmark runs, Jan-v3.5-4B-Q4_K_XL reaches up to 75.3 tok/s and averages 73.4 tok/s, with the fastest results on NVIDIA GeForce RTX 3060 Laptop GPU.
- How much memory does Jan-v3.5-4B-Q4_K_XL need?
- The leanest observed configuration used about 7.1 GB of memory (quantizations tested: Q4_K).
- Which tools have been used to run Jan-v3.5-4B-Q4_K_XL?
- Benchmarks were submitted using Jan. Results are community-contributed and updated as new runs arrive.