AMD-Pro-BenchCommunity

13,812 benchmarks in 32 GPU combinations documented

Llama-3.2-3B-Instruct on AMD GPUs

28 results · 6 sources · 31 measurements

Explore all results

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

Documented results

24 selected tests
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed
1,133.2 tok/sInput + output · mean · per request
——1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.5-v8-16k+
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
4096 / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-24
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
7,533.54 tok/sPrefill · mean · per request
—1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
Vulkan
Harness
llama-bench
Prompt / output
512 / 0 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MWeights ↗Token speed——
56.14 tok/sGeneration · per request
1 × W7700Single GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Raspberry Pi OS
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
Not reported
Full record →
View
ReportLocalMaxxing		Llama-3.2-3B-InstructQ4_K_MWeights ↗Token speed——
105.7 tok/sGeneration · median · per request
1 × Ryzen AI Max+ 395Single GPU · Unified memory · 96 GB RAM
Configuration
Memory
Unified memory · 96 GB RAM
OS
Ubuntu 24.04
Engine
Loculus Core
Backend
vulkan
Harness
LocalMaxxing community speed test; lmx 0.1.56
Prompt / output
302 / 256 tokens
Context
2048
Concurrency
1
Date
2026-10-30
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MWeights ↗Token speed——
89.31 tok/sGeneration · per request
1 × W7700Single GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Ubuntu 24.04
Engine
llama.cpp
Backend
Vulkan
Harness
llama-cli
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2024-12-10
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQuant not reportedModel ref. ↗Relative performance
1.7 x vs RTX 4500 BlackwellRelative performance per dollar · Whole system
——1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Ubuntu 24.04.3 LTS
Engine
Not reported
Backend
ROCm 6.4.2
Harness
Not reported
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
Not reported
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
3,195.65 tok/sPrefill · mean · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.5-v8-16k+
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
512 / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-24
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed——
220.39 tok/sGeneration · mean · per request
1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
Vulkan
Harness
llama-bench
Prompt / output
0 / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MWeights ↗Power & energy
145 WPower draw · Whole system
——1 × W7700Single GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Raspberry Pi OS
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
Not reported
Full record →
View
ReportLocalMaxxing		Llama-3.2-3B-InstructQ4_K_MWeights ↗Token speed
2,592.5 tok/sEstimated prefill rate · median · per request
——1 × Ryzen AI Max+ 395Single GPU · Unified memory · 96 GB RAM
Configuration
Memory
Unified memory · 96 GB RAM
OS
Ubuntu 24.04
Engine
Loculus Core
Backend
vulkan
Harness
LocalMaxxing community speed test; lmx 0.1.56
Prompt / output
302 / 256 tokens
Context
2048
Concurrency
1
Date
2026-10-30
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
2,719.87 tok/sPrefill · mean · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.5-v8-16k+
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
4096 / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-24
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
5,614.69 tok/sPrefill · mean · per request
—1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
Vulkan
Harness
llama-bench
Prompt / output
512 / 0 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed——
77.54 tok/sGeneration · mean · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.5-v8-16k+
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
? / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-24
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed——
219.8 tok/sGeneration · mean · per request
1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
Vulkan
Harness
llama-bench
Prompt / output
0 / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed
2,077.66 tok/sInput + output · mean · per request
——1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.0-6-generic
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
4096 / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-25
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
5,589.02 tok/sPrefill · mean · per request
—1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
Vulkan
Harness
llama-bench
Prompt / output
512 / 0 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
4,050.97 tok/sPrefill · mean · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.0-6-generic
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
512 / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-25
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed——
219.84 tok/sGeneration · mean · per request
1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
Vulkan
Harness
llama-bench
Prompt / output
0 / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
3,570.23 tok/sPrefill · mean · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.0-6-generic
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
4096 / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-25
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
7,361.14 tok/sPrefill · mean · per request
—1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
ROCm / HIP
Harness
llama-bench
Prompt / output
512 / 0 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed——
195.91 tok/sGeneration · mean · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.0-6-generic
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
? / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-25
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed——
178.56 tok/sGeneration · mean · per request
1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
ROCm / HIP
Harness
llama-bench
Prompt / output
0 / 128 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
View
ReportGeneric LLM		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Power & energy
281 WPower draw · Unknown
——1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux; kernel 6.17.5-v8-16k+
Engine
llama.cpp
Backend
Vulkan
Harness
Not reported
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2025-11-24
Full record →
View
Reportllama-bench		Llama-3.2-3B-InstructQ4_K_MModel ref. ↗Token speed—
7,373.63 tok/sPrefill · mean · per request
—1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPU
Configuration
Memory
16 GB VRAM/GPU
OS
Fedora 44 / kernel 7.1.3-201.fc44.x86_64
Engine
llama.cpp
Backend
ROCm / HIP
Harness
llama-bench
Prompt / output
512 / 0 tokens
Context
Not reported
Concurrency
Not reported
Date
2026-07-15
Full record →
Quantizations and software

Quantizations: Q4_K_M
Engines: llama.cpp, Loculus Core

Explore all 28 results →

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

GPU benchmark pages: W7700 · R9700 · RX 9070 XT · Ryzen AI Max+ 395