Qwen3.8-Flash-Next

180BChatby Qwen (Alibaba) Start a discussion

Open-weight vision-language model from Qwen (Alibaba), released 2026-08-24, 180.0B total, a bespoke open-weights licence. The Flash tier of the Qwen3.8 generation, roughly 180B total parameters, aimed at low-cost high-throughput serving.

0

Quantized builds

41

Benchmark records

6

Discussions

Quantized builds

Community and official quantized releases; "linked records" counts only benchmarks tied to that exact release.

No quantized builds cataloged yet.

Benchmark data

41 configurations

HardwareFrameworkQuantDecodeVRAM usedSamplesEvidence
NVIDIA RTX 5070 TiStrataQ4_K_M8.2 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 TiStrataIQ3_XXS14.2 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 TiStrataIQ3_S12.3 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 TiStrataIQ4_XS9.2 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppQ4_K_M6.4 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppIQ3_XXS11.1 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppIQ4_XS7.2 tok/s16 GB1L0 Self-reported
NVIDIA RTX 3090StrataQ4_K_M8.6 tok/s24 GB1L0 Self-reported
NVIDIA RTX 3090StrataIQ3_XXS14.8 tok/s24 GB1L0 Self-reported
NVIDIA RTX 3090StrataIQ3_S12.8 tok/s24 GB1L0 Self-reported
NVIDIA RTX 3090StrataIQ4_XS9.6 tok/s24 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppQ4_K_M6.7 tok/s24 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppIQ3_XXS11.6 tok/s24 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppIQ4_XS7.5 tok/s24 GB1L0 Self-reported
NVIDIA RTX 5090StrataQ4_K_M16.4 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090StrataIQ3_XXS28.4 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090StrataIQ3_S24.6 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090StrataIQ4_XS18.4 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppQ4_K_M12.8 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppIQ3_XXS22.1 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppIQ3_S19.2 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppIQ4_XS14.4 tok/s32 GB1L0 Self-reported
NVIDIA RTX 4090StrataQ4_K_M9.2 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090StrataIQ3_XXS15.9 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090StrataIQ3_S13.8 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090StrataIQ4_XS10.4 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppQ4_K_M7.2 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppIQ3_XXS12.4 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppIQ3_S10.8 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppIQ4_XS8.1 tok/s24 GB1L0 Self-reported
Apple M5 Maxllama.cppIQ3_S6.6 tok/s73.4 GB1L0 Self-reported
Apple M4 Maxllama.cppIQ3_S5.8 tok/s73.4 GB1L0 Self-reported
NVIDIA RTX 3060 12GBStrataIQ2_XXS7.1 tok/s12 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppIQ3_S10 tok/s24 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppIQ3_S9.6 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5090vLLMIQ3_S20.9 tok/s32 GB1L0 Self-reported
NVIDIA RTX 4090StrataIQ2_XS16.6 tok/s24 GB1L0 Self-reported
Intel Arc B580StrataIQ2_XS45 tok/s5.5 GB1L1 Reproduced
NVIDIA RTX 5070StrataIQ2_XS79 tok/s7.5 GB1L1 Reproduced
NVIDIA RTX 3060 12GBStrataIQ1_S70 tok/s6.5 GB1L1 Reproduced
4x NVIDIA Tesla P40llama.cppUD-Q3_K_XL21.6 tok/s—1L1 Reproduced

41 of these records are not linked to a specific quantized build; the full set is on the search page.

Community discussion

View all discussions