Qwen3.5-35B-A3B on AMD GPUs

117 results · 13 sources · 160 measurements

Grouped by reported model identity. Exact weights, quantizations and software can differ. How results are reviewed

Documented results

24 selected tests


View Reportllm-bench.io	Qwen3.5-35B-A3BQuant not reportedModel ref. ↗	llm-bench.io QualityFour sequential scenarios: agent workflow, code generation, role play, research	71.81 / 100LLM judge · Mean	—	31.97 tok/sGeneration · mean · per request	GPU count unknownUnknown · 24 GB VRAM/GPU Configuration Memory 24 GB VRAM/GPU OS Windows Engine LM Studio Backend Not reported Harness llm-benchmark 0.4.20+97 Prompt / output ? / ? tokens Context 16384 Concurrency 1 Date 2026-04-18 Full record →
View ReportCustom evaluation	Qwen3.5-35B-A3BGPTQ Int4 (MoE experts; BF16 other weights)Weights ↗Conflicting evidence	HumanEvalopenai/openai_humaneval test: first30 tasks1 concurrent request	50 %pass@1 · 15 / 30 · Executable tests	—	—	2 × R9700Tensor parallel · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS EndeavourOS (Arch Linux) Engine SGLang Backend ROCm Harness scripts/eval/eval_and_chart.py Prompt / output ? / ? tokens Context 32768 Concurrency 1 Date 2026-04-18 Full record →
View ReportLocalMaxxing	Qwen3.5-35B-A3BQ4_K_MModel ref. ↗	Token speed	—	742 tok/sPrefill · median · unknown	74.6 tok/sGeneration · median · unknown	1 × Ryzen AI Max+ 395Single GPU · Unified memory Configuration Memory Unified memory OS Ubuntu 24.04 Engine llama.cpp Backend vulkan Harness LocalMaxxing community speed test; exact client where reported Prompt / output 1185 / 512 tokens Context 8192 Concurrency Not reported Date 2026-08-12 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗	Token speed	—	—	153.4 tok/sGeneration · mean · per request	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.3 Harness Not reported Prompt / output ? / 128 tokens Context Not reported Concurrency Not reported Date 2026-03-27 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗	Token speed	—	—	155.2 tok/sGeneration · mean · per request	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.3 Harness Not reported Prompt / output ? / 128 tokens Context Not reported Concurrency Not reported Date 2026-03-28 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗	Token speed	—	—	156 tok/sGeneration · mean · per request	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.4 Harness Not reported Prompt / output ? / 128 tokens Context Not reported Concurrency Not reported Date 2026-04-10 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗Conflicting evidence	Token speed	—	—	116.4 tok/sGeneration · mean · per request	2 × R9700Layer split · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.4 Harness Not reported Prompt / output ? / 128 tokens Context Not reported Concurrency Not reported Date 2026-04-10 Full record →
View Reportllama-bench	Qwen3.5-35B-A3BQ4_KModel ref. ↗	Token speed	—	548.78 tok/sPrefill · per request	—	1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU Configuration Memory 16 GB VRAM/GPU OS Windows Engine LLAMA-DX / llama.cpp fork Backend Vulkan Harness llama-bench Prompt / output 512 / 0 tokens Context Not reported Concurrency Not reported Date Not reported Full record →
View ReportLocalMaxxing	Qwen3.5-35B-A3BQ3_K_MModel ref. ↗	Token speed	—	—	108.42 tok/sGeneration · peak · unknown	3 × R9700Multi-GPU · split unknown · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Arch Linux Engine llama.cpp Backend vulkan Harness LocalMaxxing community speed test; exact client where reported Prompt / output 531 / 256 tokens Context 787 Concurrency Not reported Date 2026-05-07 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BQuant not reportedModel ref. ↗	Token speed	—	—	22.83 tok/sGeneration · per request	2 × R9700Tensor parallel · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine vLLM Backend ROCm / HIP Harness Not reported Prompt / output 1036 / 128 tokens Context 131072 Concurrency 1 Date 2026-04-09 Full record →
View Reportllama-bench	Qwen3.5-35B-A3BQ4_0 custom mixWeights ↗	Token speed	—	2,900 tok/sPrefill · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Windows (version unreported) Engine llama.cpp Backend Vulkan Harness llama-bench Prompt / output 512 / ? tokens Context Not reported Concurrency 1 Date 2026-02-26 Full record →
View ReportLocalMaxxing	Qwen3.5-35B-A3BMQ4Model ref. ↗	Token speed	—	—	154.62 tok/sGeneration · unknown	1 × RX 7900 XTXSingle GPU · 24 GB VRAM/GPU Configuration Memory 24 GB VRAM/GPU OS Not reported Engine hipfire Backend Not reported Harness LocalMaxxing community speed test; exact client where reported Prompt / output 220 / 120 tokens Context 4096 Concurrency Not reported Date 2026-04-26 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BQuant not reportedModel ref. ↗	Token speed	—	—	~120 tok/sGeneration · unknown	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Windows (LM Studio); WSL2 agent workflow Engine LM Studio / llama.cpp Backend Not reported Harness Not reported Prompt / output ? / ? tokens Context Not reported Concurrency Not reported Date 2026-03-13 Full record →
View Reportllm-bench.io	Qwen3.5-35B-A3BQuant not reportedModel ref. ↗	llm-bench.io QualityFour sequential scenarios: agent workflow, code generation, role play, research	71.66 / 100LLM judge · Mean	—	29.23 tok/sGeneration · mean · per request	GPU count unknownUnknown · 24 GB VRAM/GPU Configuration Memory 24 GB VRAM/GPU OS Windows Engine LM Studio Backend Not reported Harness llm-benchmark 0.4.15+90 Prompt / output ? / ? tokens Context 16384 Concurrency 1 Date 2026-04-17 Full record →
View ReportLocalMaxxing	Qwen3.5-35B-A3BQ4_K_MModel ref. ↗	Token speed	—	739.9 tok/sPrefill · median · unknown	74.7 tok/sGeneration · median · unknown	1 × Ryzen AI Max+ 395Single GPU · Unified memory Configuration Memory Unified memory OS Ubuntu 24.04 Engine llama.cpp Backend vulkan Harness LocalMaxxing community speed test; exact client where reported Prompt / output 1185 / 512 tokens Context 8192 Concurrency Not reported Date 2026-08-12 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗	Token speed	—	—	153.1 tok/sGeneration · mean · per request	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.3 Harness Not reported Prompt / output ? / 128 tokens Context Not reported Concurrency Not reported Date 2026-03-27 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗	Token speed	—	—	114.2 tok/sGeneration · mean · per request	2 × R9700Row split · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.3 Harness Not reported Prompt / output ? / 128 tokens Context Not reported Concurrency Not reported Date 2026-03-28 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗	Token speed	—	—	115.4 tok/sGeneration · mean · per request	2 × R9700Row split · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.4 Harness Not reported Prompt / output ? / 128 tokens Context Not reported Concurrency Not reported Date 2026-04-10 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BUD-Q4_K_XLModel ref. ↗Conflicting evidence	Token speed	—	—	116.8 tok/sGeneration · mean · per request	2 × R9700Layer split · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS; Linux 6.19.8-061908-generic Engine llama.cpp Backend Vulkan / Mesa RADV 26.0.4 Harness Not reported Prompt / output ? / 512 tokens Context Not reported Concurrency Not reported Date 2026-04-10 Full record →
View Reportllama-bench	Qwen3.5-35B-A3BQ4_KModel ref. ↗	Token speed	—	—	50.64 tok/sGeneration · per request	1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU Configuration Memory 16 GB VRAM/GPU OS Windows Engine LLAMA-DX / llama.cpp fork Backend Vulkan Harness llama-bench Prompt / output 0 / 128 tokens Context Not reported Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BQuant not reportedModel ref. ↗	Token speed	—	~4,180 tok/sPrefill · range lower · aggregate	—	2 × R9700Tensor parallel · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine vLLM Backend ROCm / HIP Harness Not reported Prompt / output ? / ? tokens Context Not reported Concurrency Not reported Date 2026-04-09 Full record →
View Reportllama-bench	Qwen3.5-35B-A3BQ4_0 custom mixWeights ↗	Token speed	—	2,815 tok/sPrefill · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Windows (version unreported) Engine llama.cpp Backend Vulkan Harness llama-bench Prompt / output 1024 / ? tokens Context Not reported Concurrency 1 Date 2026-02-26 Full record →
View ReportLocalMaxxing	Qwen3.5-35B-A3BMQ4Model ref. ↗	Token speed	—	—	140.59 tok/sGeneration · unknown	1 × RX 7900 XTXSingle GPU · 24 GB VRAM/GPU Configuration Memory 24 GB VRAM/GPU OS Not reported Engine hipfire Backend Not reported Harness LocalMaxxing community speed test; exact client where reported Prompt / output 220 / 120 tokens Context 4096 Concurrency Not reported Date 2026-04-26 Full record →
View ReportGeneric LLM	Qwen3.5-35B-A3BQuant not reportedModel ref. ↗	Latency10,000 input	~4,400 msPrefill time · Unknown	—	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Windows (LM Studio); WSL2 agent workflow Engine LM Studio / llama.cpp Backend Not reported Harness Not reported Prompt / output 10000 / ? tokens Context Not reported Concurrency Not reported Date 2026-03-13 Full record →

Quantizations and software

Quantizations: GPTQ Int4 (MoE experts; BF16 other weights), MQ4, Q3_K_M, Q4_0 custom mix, Q4_K, Q4_K_M, Q8_0, UD-Q4_K_XL
Engines: hipfire, LLAMA-DX / llama.cpp fork, llama.cpp, LM Studio, LM Studio / llama.cpp, SGLang, vLLM

Explore all 117 results →

This is a selection across sources, not a ranking or a matched experiment. Reading benchmark reports

GPU benchmark pages: R9700 · RX 9070 XT · RX 7900 XTX · Ryzen AI Max+ 395