RX 9070 LLM benchmarks
96 results · 143 measurements · 12 sources
Device specifications
| Variant | Memory | Memory type | Bus | Bandwidth | FP16 matrix | INT8 matrix |
|---|---|---|---|---|---|---|
| RX 9070 ↗ | 16 GB | GDDR6 | 256-bit | 640 GB/s | 145 TFLOPS | 289 TOPS |
Manufacturer specifications. Shared system RAM is not dedicated VRAM. Unreported values remain —; measured LLM results retain their reported configuration.
Selected reported tests
24 of 96 resultsSource-reported results, selected across reports by documentation quality. Match configurations before comparing values.
| View ReportGeneric LLM | Qwen3.6-35B-A3B | HumanEvalFull 164 problems | 59.15 %pass@1 · 97 / 164 · Executable tests 38 tok/sThroughput · aggregate | — | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
|---|---|---|---|---|---|---|
| View ReportGeneric LLM | Qwen3.5-9B | HumanEvalFull 164 problems | 67.07 %pass@1 · n = 164 · Executable tests | — | 37.98 tok/sGeneration · mean · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.6-27B | HumanEvalFull164problems | 78.05 %pass@1 · 128 / 164 · Executable tests 16.8 tok/sThroughput · aggregate | — | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportLocalMaxxing | Qwen3.6-35B-A3B | Token speed | — | 399 tok/sPrefill · unknown | 47 tok/sGeneration · unknown | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Qwen3.5-9BConflicting evidence | Token speed | — | 2,115.89 tok/sPrefill · mean · per request | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3-8B (checkpoint unspecified) | Token speed | — | 2,511.81 tok/sPrefill · mean · per request | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | gemma-4-26B-A4B-it | Token speed | — | 3,868 tok/sPrefill · per request | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | DeepSeek-R1-Distill-Llama-8B | Token speed | — | — | 70.65 tok/sGeneration · mean · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-2-7B | Token speed | — | 2,361.1 tok/sPrefill · mean · per request | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Llama-2-13b-chat | Procyon | — | — | 59.67 tok/sGeneration · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | — | 22.1 tok/sGeneration · per request | 2 × RX 9070Multi-GPU · split unknown · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Llama-3.1-8B-Instruct | Procyon | 1,787 pointsBenchmark score · Per request | — | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.5-9B | HumanEvalFull 164 problems | 67.07 %pass@1 · n = 164 · Executable tests | — | 39.8 tok/sGeneration · mean · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportCoding session | Qwen3.6-27BConflicting evidence | PinchBenchFive-task smoke / sanity,calendar,summary,weather,csv_iris_summary1 concurrent request | 4.46 / 5LLM judge | — | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportLocalMaxxing | Qwen3.6-35B-A3B | Token speed | — | — | 34.2 tok/sGeneration · unknown | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Qwen3.5-9BConflicting evidence | Token speed | — | — | 86.45 tok/sGeneration · mean · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-3-8B (checkpoint unspecified) | Token speed | — | — | 102.31 tok/sGeneration · mean · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | gemma-4-26B-A4B-it | Token speed | — | — | 157.2 tok/sGeneration · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | DeepSeek-R1-Distill-Llama-8B | Token speed | — | 3,478.42 tok/sPrefill · mean · per request | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View Reportllama-bench | Llama-2-7B | Token speed | — | — | 99.39 tok/sGeneration · mean · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Llama-3.1-8B-Instruct | Procyon | — | — | 87.59 tok/sGeneration · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | — | 21 tok/sGeneration · per request | 2 × RX 9070Multi-GPU · split unknown · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.5-9B | HumanEvalFull 164 problems | 66.46 %pass@1 · n = 164 · Executable tests | — | 38.76 tok/sGeneration · mean · per request | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportCoding session | Qwen3.6-27B | PinchBenchFive-task smoke1 concurrent request | 100 %LLM judge · n = 1 | — | — | 1 × RX 9070Single GPU · 16 GB VRAM/GPUConfiguration
|
Models and software in this collection
Model families: DeepSeek R1, GPT-OSS, Gemma 4, Granite 3.0, Llama 2, Llama 3, Llama 3.1, Mistral, Phi 3.5, Qwen3, Qwen3.5, Qwen3.6, Qwen3.8
Engines: ONNX Runtime, llama.cpp, llama.cpp TurboQuant fork
Methodology and evidence standards · Download the research dataset