Qwen3.6-27B

27BChatby Qwen (Alibaba) Start a discussion

Qwen3.6's 27B dense model (released 2026-04-22), Apache-2.0. Positioned for coding and agentic work at the 27B scale, with a vision encoder and hybrid attention (Gated DeltaNet plus periodic full attention).

0

Quantized builds

40

Benchmark records

1

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

39 configurations

HardwareFrameworkQuantDecodeVRAM usedSamplesEvidence
NVIDIA RTX Sparkllama.cppFP165.5 tok/s107 GB1L2 Cross-framework verified
NVIDIA RTX Sparkllama.cppQ5_K_M8.4 tok/s54.83 GB1L2 Cross-framework verified
NVIDIA RTX Sparkllama.cppQ8_07.7 tok/s67.3 GB1L2 Cross-framework verified
NVIDIA RTX SparkOllamaQ5_K_M8.8 tok/s54.83 GB1L2 Cross-framework verified
NVIDIA RTX SparkOllamaQ8_06.7 tok/s67.3 GB1L2 Cross-framework verified
NVIDIA RTX 3090vLLMIQ1_S96.1 tok/s22.8 GB1L2 Cross-framework verified
AMD Ryzen AI Max+ 395llama.cppIQ3_XS27.7 tok/s37.25 GB1L2 Cross-framework verified
AMD Ryzen AI Max+ 395OllamaFP1612 tok/s107 GB1L2 Cross-framework verified
Apple M5 Maxllama.cppIQ1_S83.4 tok/s25.91 GB1L2 Cross-framework verified
Apple M5 MaxOllamaIQ4_XS48.6 tok/s44.06 GB1L2 Cross-framework verified
Apple M5 Ultrallama.cppIQ3_XS91.1 tok/s37.25 GB1L2 Cross-framework verified
Apple M5 Ultrallama.cppQ4_K_S81.1 tok/s46.33 GB1L2 Cross-framework verified
Apple M5 UltraOllamaIQ3_S85.9 tok/s39.52 GB1L2 Cross-framework verified
Apple M5 UltraOllamaQ6_K53.3 tok/s58.8 GB1L2 Cross-framework verified
NVIDIA H100 80GB HBM3vLLMQ8_0234.2 tok/s67.3 GB1L2 Cross-framework verified
NVIDIA H100 80GB HBM3StrataQ3_K_S335.3 tok/s38.96 GB1L2 Cross-framework verified
NVIDIA H100 80GB HBM3StrataQ8_0297.9 tok/s67.3 GB1L2 Cross-framework verified
NVIDIA H20vLLMQ3_K_S48.5 tok/s38.96 GB1L2 Cross-framework verified
NVIDIA H20OllamaQ4_K_M33 tok/s48.59 GB1L2 Cross-framework verified
NVIDIA RTX PRO 6000 Blackwell 96GBllama.cppIQ1_S267.3 tok/s25.91 GB1L2 Cross-framework verified
NVIDIA RTX PRO 6000 Blackwell 96GBllama.cppQ8_0105.8 tok/s67.3 GB1L2 Cross-framework verified
NVIDIA RTX PRO 6000 Blackwell 96GBvLLMIQ3_S220.1 tok/s39.52 GB1L2 Cross-framework verified
Apple M2 Ultrallama.cppQ4_K_M67.4 tok/s48.59 GB1L2 Cross-framework verified
Apple M2 UltraOllamaQ3_K_L80.8 tok/s42.92 GB1L2 Cross-framework verified
NVIDIA RTX Sparkllama.cppIQ3_S13.8 tok/s39.52 GB1L2 Cross-framework verified
NVIDIA RTX Sparkllama.cppQ4_K_M10.9 tok/s48.59 GB1L2 Cross-framework verified
NVIDIA RTX SparkOllamaQ4_K_M10.4 tok/s48.59 GB2L2 Cross-framework verified
NVIDIA RTX 3090OllamaIQ1_S76.6 tok/s22.8 GB1L2 Cross-framework verified
NVIDIA RTX 5090OllamaIQ2_M234.2 tok/s30.4 GB1L2 Cross-framework verified
Apple M3 MaxOllamaIQ2_XS109.2 tok/s29.32 GB1L2 Cross-framework verified
NVIDIA RTX 5090llama.cppIQ1_S273.1 tok/s25.91 GB1L2 Cross-framework verified
Apple M4 Maxllama.cppIQ2_M189.9 tok/s31.58 GB1L2 Cross-framework verified
Apple M4 Maxllama.cppIQ2_XS191.2 tok/s29.32 GB1L2 Cross-framework verified
Apple M4 MaxvLLMQ3_K_M174.2 tok/s40.66 GB1L2 Cross-framework verified
Apple M4 MaxOllamaQ6_K99.5 tok/s58.8 GB1L2 Cross-framework verified
NVIDIA RTX 3090llama.cppQ4_K_M57 tok/s23.44 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppUD-Q3_K_XL36.7 tok/s20.11 GB1L0 Self-reported
NVIDIA RTX 3090llama.cppQ5_K_S46.6 tok/s21.53 GB1L0 Self-reported
NVIDIA RTX 3090 Tillama.cppQ5_K_S50.8 tok/s22.47 GB1L0 Self-reported

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

Community discussion

View all discussions