Appearance
Compare Qwen3.8-Flash-Next token generation and prompt processing across quantizations on Olares One, NVIDIA DGX Spark and Mac Studio M5 Max.
20 prompts, medium reasoning, 3,072-token output limit
One request at a time. Hardware and Strata settings differ between systems. Q2_0 has speed results only, with no IFBench evaluation.
Token generation, single request
Higher is betterScroll horizontally to view all metrics.
| Machine | Configuration | Token generation | Time to first token |
|---|---|---|---|
| Olares OneStrata, Q2_0 ↗ | Strata, Q2_0 ↗ | 166 tok/s | 0.29 s |
| Olares OneStrata, IQ2_XS ↗ | Strata, IQ2_XS ↗ | 162 tok/s | 0.32 s |
| Olares OneStrata, IQ3_XXS ↗ | Strata, IQ3_XXS ↗ | 133 tok/s | 0.34 s |
| Mac Studio M5 MaxStrata, Q2_0 ↗ | Strata, Q2_0 ↗ | 130 tok/s | 0.25 s |
| Olares OneStrata, IQ3_S ↗ | Strata, IQ3_S ↗ | 113 tok/s | 0.39 s |
| Olares OneStrata, UD-IQ4_XS ↗ | Strata, UD-IQ4_XS ↗ | 92 tok/s | 0.43 s |
| DGX SparkStrata, Q2_0 ↗ | Strata, Q2_0 ↗ | 80 tok/s | 0.53 s |
| DGX SparkStrata, IQ3_XXS ↗ | Strata, IQ3_XXS ↗ | 68 tok/s | 0.60 s |
| DGX SparkStrata, IQ3_S ↗ | Strata, IQ3_S ↗ | 59 tok/s | 0.64 s |
| DGX SparkStrata, UD-IQ4_XS ↗ | Strata, UD-IQ4_XS ↗ | 52 tok/s | 0.64 s |
| Mac Studio M5 MaxStrata, IQ3_XXS, no MTP ↗ | Strata, IQ3_XXS, no MTP ↗ | 47 tok/s | 1.01 s |
| Mac Studio M5 MaxStrata, IQ3_S, no MTP ↗ | Strata, IQ3_S, no MTP ↗ | 41 tok/s | 1.35 s |
Client-observed prompt processing speed, median of three runs per prompt length
Calculated as prompt tokens divided by time to first token. This differs from the engine-reported metric on the Qwen3.8-27B page. Use the same metric when comparing models.
Prompt length in tokens. Speed in tokens per second.
Scroll horizontally to compare all prompt lengths.
| Machine | Configuration | 1k | 4k | 16k | 32k | 64k |
|---|---|---|---|---|---|---|
| Olares OneStrata, Q2_0 ↗ | Strata, Q2_0 ↗ | 1,257 | 2,536 | 3,261 | 3,321 | 3,312 |
| Olares OneStrata, IQ3_S ↗ | Strata, IQ3_S ↗ | 910 | 2,235 | 2,522 | 2,603 | 2,447 |
| Olares OneStrata, IQ3_XXS ↗ | Strata, IQ3_XXS ↗ | 1,034 | 2,349 | 2,374 | 2,285 | 2,359 |
| Olares OneStrata, IQ2_XS ↗ | Strata, IQ2_XS ↗ | 1,181 | 1,834 | 2,268 | 2,295 | 2,285 |
| Mac Studio M5 MaxStrata, Q2_0 ↗ | Strata, Q2_0 ↗ | 1,998 | 2,066 | 2,128 | 2,105 | 2,074 |
| DGX SparkStrata, UD-IQ4_XS ↗ | Strata, UD-IQ4_XS ↗ | 667 | 1,667 | 2,095 | 2,183 | 2,169 |
| DGX SparkStrata, IQ3_XXS ↗ | Strata, IQ3_XXS ↗ | 694 | 1,693 | 2,006 | 1,885 | 2,058 |
| DGX SparkStrata, Q2_0 ↗ | Strata, Q2_0 ↗ | 766 | 1,477 | 1,892 | 1,880 | 1,893 |
| Olares OneStrata, UD-IQ4_XS ↗ | Strata, UD-IQ4_XS ↗ | 782 | 1,219 | 1,872 | 1,850 | 1,867 |
| DGX SparkStrata, IQ3_S ↗ | Strata, IQ3_S ↗ | 620 | 1,554 | 1,865 | 1,933 | 1,966 |
| Mac Studio M5 MaxStrata, IQ3_XXS, no MTP ↗ | Strata, IQ3_XXS, no MTP ↗ | 565 | 613 | 624 | 616 | 590 |
| Mac Studio M5 MaxStrata, IQ3_S, no MTP ↗ | Strata, IQ3_S, no MTP ↗ | 431 | 455 | 463 | 463 | 449 |