AMD-Pro-BenchCommunity

13,812 benchmarks in 32 GPU combinations documented

Qwen3.8-27B-PARO on R9700

109 results · 2 sources · 112 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
ReportCustom evaluation		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗GSM8K500 questions
97.4 %Accuracy · 487 / 500 · Reported correctness
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
GSM8K evaluation (client unreported)
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2026-08-31
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
2,702 tok/sInput / client TTFT · per request
21.03 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
2042 / 512 tokens
Context
131072
Concurrency
1
Date
2026-09-12
Full record →
View
ReportCustom evaluation		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗GSM8K500 questions
97.6 %Accuracy · 488 / 500 · Reported correctness
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
GSM8K evaluation (client unreported)
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2026-09-02
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
2,909.7 tok/sInput / client TTFT · per request
20.1 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
32193 / 512 tokens
Context
131072
Concurrency
1
Date
2026-09-12
Full record →
View
ReportCustom evaluation		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗GSM8K500 questions
97.4 %Accuracy · 487 / 500 · Reported correctness
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
GSM8K evaluation (client unreported)
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2026-09-03
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
4,759.8 tok/sInput / client TTFT · per request
19.46 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
60355 / 512 tokens
Context
131072
Concurrency
1
Date
2026-09-12
Full record →
View
ReportCustom evaluation		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗WikiText96 x 500-character chunks
0.0195 natsScore · KL divergence on reference top-K · Lower is better
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
kld.py
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
3,134.8 tok/sInput / client TTFT · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
2042 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportCustom evaluation		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗WikiText96 x 500-character chunks
0.0285 natsScore · KL divergence on reference top-K · Lower is better
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
kld.py
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
2,818.2 tok/sInput / client TTFT · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
32193 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportCustom evaluation		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗WikiText96 x 500-character chunks
91.5 %Score · n = 100 · Top-1 token agreement
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
kld.py
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
Not reported
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
4,493.7 tok/sInput / client TTFT · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
60355 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportBetterBench		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗Token speed——
169.8 tok/sGeneration · median · per request
2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
BetterBench
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
1
Date
2026-08-31
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
3,542.2 tok/sInput / client TTFT · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
2042 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportBetterBench		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗Token speed
466.9 tok/sThroughput · aggregate
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
BetterBench
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
8
Date
2026-08-31
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
3,238.3 tok/sInput / client TTFT · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
32193 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportBetterBench		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗Token speed
476.2 tok/sThroughput · aggregate
——2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
BetterBench
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
16
Date
2026-08-31
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed—
5,128.1 tok/sInput / client TTFT · per request
—1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab.py
Prompt / output
60355 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportBetterBench		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗Token speed—
3,128 tok/sPrefill · per request
—2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
BetterBench
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
1
Date
2026-08-31
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed——
61.88 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab2.py
Prompt / output
3495 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportBetterBench		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗Token speed—
3,083 tok/sPrefill · per request
—2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
BetterBench
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
1
Date
2026-08-31
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed——
86.51 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab2.py
Prompt / output
3590 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
View
ReportBetterBench		Qwen3.8-27B-PAROINT4 ParoQuant W4A8Weights ↗Token speed—
3,120 tok/sPrefill · per request
—2 × R9700Tensor parallel · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
Linux (inferred; host distribution unreported) (inferred)
Engine
vLLM Radiance
Backend
ROCm / HIP
Harness
BetterBench
Prompt / output
? / ? tokens
Context
Not reported
Concurrency
1
Date
2026-08-31
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-PAROMXFP4 W4A8Weights ↗Token speed——
57.95 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
vLLM
Backend
ROCm / HIP
Harness
vllm_ab2.py
Prompt / output
55254 / 512 tokens
Context
65536
Concurrency
1
Date
2026-09-12
Full record →
Quantizations and software

Quantizations: INT4 ParoQuant W4A8, MXFP4 W4A8
Engines: vLLM, vLLM Radiance

Explore all 109 results →

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

GPU benchmark pages: R9700