Qwen3.5-35B-A3B

36BChatby Qwen (Alibaba) Start a discussion

Open-weight vision-language model from Qwen (Alibaba), released 2026-02-24, 36.0B total with ~3B active, Apache-2.0. MoE with 36B total and about 3B active parameters, for setups that are VRAM-bound but want a larger knowledge pool.

0

Quantized builds

25

Benchmark records

Quantized builds

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

No quantized builds cataloged yet.

Benchmark data

25 configurations

HardwareFrameworkQuantDecodeVRAM usedSamplesEvidence
Apple M3 MaxOllamaIQ2_M78.9 tok/s34.2 GB1L2 Cross-framework verified
Apple M3 MaxStrataIQ2_M121.5 tok/s34.2 GB1L2 Cross-framework verified
NVIDIA RTX 5090vLLMIQ1_S227.8 tok/s30.4 GB1L2 Cross-framework verified
Apple M4 MaxvLLMIQ4_XS115.3 tok/s58.04 GB1L2 Cross-framework verified
Apple M4 MaxvLLMQ4_K_S115.9 tok/s61.07 GB1L2 Cross-framework verified
Apple M4 MaxOllamaIQ2_XS138 tok/s38.39 GB1L2 Cross-framework verified
NVIDIA RTX 4080 Superllama.cppQ6_K18.2 tok/s16 GB1L0 Self-reported
NVIDIA RTX 4080 Superllama.cppQ5_K_S23.6 tok/s16 GB1L0 Self-reported
NVIDIA RTX 4080 Superllama.cppIQ3_S39.3 tok/s15.1 GB1L0 Self-reported
NVIDIA RTX 4080 Superllama.cppIQ4_XS29.6 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppQ6_K22.1 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppQ5_K_S28.8 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppIQ3_S47.9 tok/s15.1 GB1L0 Self-reported
NVIDIA RTX 5070 Tillama.cppIQ4_XS36 tok/s16 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppQ6_K44.2 tok/s32 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppQ5_K_S57.5 tok/s24.8 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppIQ3_S95.7 tok/s15.1 GB1L0 Self-reported
NVIDIA RTX 5090llama.cppIQ4_XS72 tok/s19.9 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppQ6_K24.9 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppQ5_K_S32.4 tok/s24 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppIQ3_S53.9 tok/s15.1 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppIQ4_XS40.5 tok/s19.9 GB1L0 Self-reported
Apple M5 Maxllama.cppQ4_K_M21.9 tok/s22.4 GB1L0 Self-reported
NVIDIA RTX 5090vLLMQ4_K_M69.6 tok/s22.4 GB1L0 Self-reported
NVIDIA RTX 4090llama.cppQ4_K_M35.9 tok/s22.4 GB1L0 Self-reported

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

Community discussion

No discussions yet — start the first one.

Start a discussion