← Overview
qwen3-coder-next
- Quantization
- Q6
- Reasoning
- off
- Profile
- quick
- Suite
- v2b
- Date
- 2026-09-23 – 2026-09-24
- Hardware
- NVIDIA DGX Spark, 128 GB RAM
- Inference software
- vLLM 0.26.0
- Measured
- directly at the inference server
- Overall
- 65.1
Categories
| Category | Score |
|---|
| Knowledge & reasoning | 71.3 |
| Math | 61.6 |
| Instruction following | 45.9 |
| Code | 70.5 |
| German | 76.0 |
Tasks
| Category | Task | Score | ± | n |
|---|
| Knowledge & reasoning |
mmlu_pro |
82.0 |
± 2.2 |
300 |
| Knowledge & reasoning |
gpqa_diamond_gen |
68.2 |
± 3.3 |
198 |
| Knowledge & reasoning |
musr_gen |
63.6 |
± 3.0 |
250 |
| Math |
omni_math_gen |
61.6 |
± 3.1 |
250 |
| Instruction following |
ifbench |
32.7 |
± 2.7 |
300 |
| Instruction following |
ifeval |
86.0 |
± 2.8 |
150 |
| Code |
livecodebench_gen |
70.5 |
± 3.2 |
200 |
| German |
mmlu_prox_de |
78.7 |
± 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] |
– |
– |
43.7 |
43.5 |
43.8 |
5 |
| Prompt processing [t/s] |
512 |
525 |
744.0 |
716.4 |
768.1 |
5 |
| Prompt processing [t/s] |
4096 |
4028 |
987.8 |
986.6 |
998.2 |
5 |
| Prompt processing [t/s] |
16384 |
16045 |
1030.1 |
1028.9 |
1033.0 |
5 |
| Time to first token [ms] |
512 |
525 |
705.6 |
683.5 |
732.0 |
5 |
| Time to first token [ms] |
4096 |
4028 |
4077.6 |
4035.2 |
4082.8 |
5 |
| Time to first token [ms] |
16384 |
16045 |
15576.0 |
15532.7 |
15593.8 |
5 |
Measured directly at the inference server.