AMD-Pro-BenchCommunity

13,812 benchmarks in 32 GPU combinations documented

Qwen3.8-27B-heretic-ara on R9700

77 results · 3 sources · 141 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		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
854.56 tok/sPrefill · per request
53.08 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
phase1_turbo_ab.py
Prompt / output
38723 / 586 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
990.8 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
llama.cpp
Backend
Vulkan
Harness
vllm_ab.py
Prompt / output
2003 / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcode_tok_s——
125.02 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
1,012.4 tok/sPrefill · per request
34.58 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
2363 / 768 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
1,013.1 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
llama.cpp
Backend
Vulkan
Harness
vllm_ab.py
Prompt / output
32193 / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcopy_tok_s——
228.05 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
900.32 tok/sPrefill · per request
31.92 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
38724 / 768 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
711.2 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
llama.cpp
Backend
Vulkan
Harness
vllm_ab.py
Prompt / output
128310 / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcode_tok_s——
123.84 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
622.48 tok/sPrefill · per request
27.03 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
128939 / 768 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed——
52.53 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
llama.cpp
Backend
Vulkan
Harness
vllm_ab2.py
Prompt / output
2690 / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcopy_tok_s——
225.06 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
990.33 tok/sPrefill · per request
34.59 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
2362 / 721 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed——
65.3 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
llama.cpp
Backend
Vulkan
Harness
vllm_ab2.py
Prompt / output
2785 / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcode_tok_s——
123.56 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
1,012.37 tok/sPrefill · per request
33.56 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
14792 / 461 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed——
44.77 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
llama.cpp
Backend
Vulkan
Harness
vllm_ab2.py
Prompt / output
41817 / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcopy_tok_s——
224.98 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
895.83 tok/sPrefill · per request
31.92 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
38724 / 573 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed——
53.63 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
llama.cpp
Backend
Vulkan
Harness
vllm_ab2.py
Prompt / output
41912 / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcode_edit——
77.93 tok/sGeneration · mean · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed—
622.36 tok/sPrefill · per request
27.03 tok/sGeneration · per request
1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
128939 / 597 tokens
Context
262144
Concurrency
1
Date
2026-09-10
Full record →
View
ReportServing benchmark		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speed
70.06 tok/sThroughput · mean · per request
——1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS / Linux 7.1.x
Engine
llama.cpp
Backend
Vulkan
Harness
vllm_conc2.py (conc_test.py method)
Prompt / output
113 / ? tokens
Context
Not reported
Concurrency
1
Date
2026-09-12
Full record →
View
ReportGeneric LLM		Qwen3.8-27B-heretic-araQ4_K_SModel ref. ↗Token speedcode_edit
76.76 tok/sThroughput · aggregate
——1 × R9700Single GPU · 32 GB VRAM/GPU
Configuration
Memory
32 GB VRAM/GPU
OS
CachyOS (Arch-based)
Engine
llama.cpp
Backend
Vulkan / Mesa RADV 26.x
Harness
Not reported
Prompt / output
? / 512 tokens
Context
262144
Concurrency
1
Date
2026-09-11
Full record →
Quantizations and software

Quantizations: Q4_K_S
Engines: llama.cpp

Explore all 77 results →

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

GPU benchmark pages: R9700