Qwen2.5-32B-Instruct on AMD GPUs

16 results · 2 sources · 22 measurements

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

Documented results

16 selected tests


View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Token speed	—	—	28.29 tok/sGeneration · per request	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine llama.cpp / llama-cpp-python Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / 149 tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MModel ref. ↗	Token speed	—	633 tok/sPrefill · mean · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output 128 / ? tokens Context 8192 Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Token speed	—	—	28.68 tok/sGeneration · per request	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine llama.cpp / llama-cpp-python Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / 151 tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MModel ref. ↗	Token speed	—	825 tok/sPrefill · mean · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output 256 / ? tokens Context 8192 Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Token speed	—	—	28.7 tok/sGeneration · per request	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine llama.cpp / llama-cpp-python Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / 151 tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MModel ref. ↗	Token speed	—	901 tok/sPrefill · mean · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output 512 / ? tokens Context 8192 Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Token speed	—	—	22.77 tok/sGeneration · per request	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine VTE Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / 144 tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MModel ref. ↗	Token speed	—	—	26.8 tok/sGeneration · mean · per request	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output ? / 128 tokens Context 8192 Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Token speed	—	—	22.62 tok/sGeneration · per request	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine VTE Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / 150 tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ5_K_MModel ref. ↗	Token speed	—	616 tok/sPrefill · mean · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output 128 / ? tokens Context 8192 Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Token speed	—	—	22.7 tok/sGeneration · per request	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine VTE Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / 150 tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ5_K_MModel ref. ↗	Token speed	—	781 tok/sPrefill · mean · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output 256 / ? tokens Context 8192 Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Latency1 concurrent request	3,449.81 msModel load time · Whole system	—	—	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine llama.cpp / llama-cpp-python Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / ? tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ5_K_MModel ref. ↗	Token speed	—	854 tok/sPrefill · mean · per request	—	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output 512 / ? tokens Context 8192 Concurrency Not reported Date Not reported Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ4_K_MWeights ↗	Latency1 concurrent request	28,337.76 msModel load time · Whole system	—	—	1 × W7900Single GPU · 48 GB VRAM/GPU Configuration Memory 48 GB VRAM/GPU OS Ubuntu 24.04.4 LTS Engine VTE Backend HIP / ROCm 7.2.1 Harness VTE Python CLI benchmark Prompt / output ? / ? tokens Context 4096 Concurrency 1 Date 2026-07-19 Full record →
View ReportGeneric LLM	Qwen2.5-32B-InstructQ5_K_MModel ref. ↗	Token speed	—	—	23.7 tok/sGeneration · mean · per request	1 × R9700Single GPU · 32 GB VRAM/GPU Configuration Memory 32 GB VRAM/GPU OS Ubuntu 24.04.4 Engine llama.cpp Backend HIP / ROCm 7.2.0 Harness Not reported Prompt / output ? / 128 tokens Context 8192 Concurrency Not reported Date Not reported Full record →

Quantizations and software

Quantizations: Q4_K_M, Q5_K_M
Engines: llama.cpp, llama.cpp / llama-cpp-python, VTE

Explore all 16 results →

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

GPU benchmark pages: W7900 · R9700