← Overview
qwen3.8-27b-mtp
- Quantization
- NVFP4
- Reasoning
- off
- Profile
- quick
- Suite
- v2b
- Date
- 2026-09-18
- Hardware
- NVIDIA DGX Spark, 128 GB RAM
- Inference software
- vLLM 0.26.0
- Measured
- directly at the inference server
- Overall
- 64.2
Categories
| Category | Score |
|---|
| Knowledge & reasoning | 69.9 |
| Math | 60.4 |
| Instruction following | 49.5 |
| Code | 65.0 |
| German | 76.3 |
Tasks
| Category | Task | Score | ± | n |
|---|
| Knowledge & reasoning |
mmlu_pro |
80.3 |
± 2.3 |
300 |
| Knowledge & reasoning |
gpqa_diamond_gen |
63.6 |
± 3.4 |
198 |
| Knowledge & reasoning |
musr_gen |
65.6 |
± 3.0 |
250 |
| Math |
omni_math_gen |
60.4 |
± 3.1 |
250 |
| Instruction following |
ifbench |
38.3 |
± 2.8 |
300 |
| Instruction following |
ifeval |
83.3 |
± 3.1 |
150 |
| Code |
livecodebench_gen |
65.0 |
± 3.4 |
200 |
| German |
mmlu_prox_de |
79.0 |
± 2.4 |
300 |
| German |
include_de_gen |
70.8 |
± 3.7 |
154 |
± is the standard error; the 95% interval spans roughly 1.96 standard errors around the score.
Speed
| Measure | Context (target) | Context (actual) | Median | p10 | p90 | n |
|---|
| Generation [t/s] |
– |
– |
15.1 |
14.3 |
15.6 |
5 |
| Prompt processing [t/s] |
512 |
529 |
1836.6 |
1822.4 |
1852.5 |
5 |
| Prompt processing [t/s] |
4096 |
4024 |
2550.6 |
2342.9 |
2556.1 |
5 |
| Prompt processing [t/s] |
16384 |
16001 |
1734.3 |
1676.5 |
1735.1 |
5 |
| Time to first token [ms] |
512 |
529 |
288.0 |
285.3 |
290.3 |
5 |
| Time to first token [ms] |
4096 |
4024 |
1577.7 |
1574.2 |
1726.3 |
5 |
| Time to first token [ms] |
16384 |
16001 |
9226.4 |
9221.8 |
9551.4 |
5 |
Measured directly at the inference server.