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 | Qwen2.5-3B-Instruct | llm-bench.io QualityFour sequential scenarios: agent workflow, code generation, role play, research | 40.9 / 100LLM judge · Mean | 3,654.68 tok/sPrefill · mean · per request | 74.25 tok/sGeneration · mean · per request | GPU count unknownUnknown · 16 GB VRAM/GPUConfiguration
|
|---|---|---|---|---|---|---|
| View Reportllama-bench | Qwen2.5-3B-Instruct | Token speed | — | 3,117.23 tok/sPrefill · mean · per request | — | 1 × RX 9060 XTSingle GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Token speed | 53.3 tok/sThroughput · per request | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | 3,999 tok/sPrefill · mean · per request | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Qwen2.5-3B-Instruct | Token speed | — | — | 79.02 tok/sGeneration · mean · per request | 1 × RX 9060 XTSingle GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Token speed | 183 tok/sThroughput · aggregate | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | 6,387 tok/sPrefill · mean · per request | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Qwen2.5-3B-Instruct | Token speed | — | 2,414.06 tok/sPrefill · mean · per request | — | 1 × RX 9060 XTSingle GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Token speed | 251 tok/sThroughput · aggregate | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | 8,124 tok/sPrefill · mean · per request | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Qwen2.5-3B-Instruct | Token speed | — | — | 70.83 tok/sGeneration · mean · per request | 1 × RX 9060 XTSingle GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Latency1 concurrent request | 60 msTime to first token · Per request | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | — | 143.4 tok/sGeneration · mean · per request | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Latency1 concurrent request | 19 msInter-token latency · Per request | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | 3,836 tok/sPrefill · mean · per request | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Latency1 concurrent request | 6,100 msLatency · Per request | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | 6,106 tok/sPrefill · mean · per request | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Latency1 concurrent request | 7,500 msLatency · Per request | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | 7,737 tok/sPrefill · mean · per request | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Power & energy1 concurrent request | 182 WPower draw · Per GPU | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | — | 137 tok/sGeneration · mean · per request | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Latency4 concurrent requests | 76 msTime to first token · Per request | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen2.5-3B-Instruct | Token speed | — | 1,325 tok/sPrefill · mean · per request | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportServing benchmark | Qwen2.5-3B-InstructConflicting evidence | Latency4 concurrent requests | 21 msInter-token latency · Per request | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
Quantizations and software
Quantizations: FP16, Q4_K_M, Q5_K_M, Q8_0
Engines: llama.cpp, vLLM
This is a selection across sources, not a ranking or a matched experiment. Reading benchmark reports
GPU benchmark pages: R9700 · RX 9060 XT · RX 7800 XT