Grouped by reported model identity. Exact weights, quantizations and software can differ. How results are reviewed
Documented results
24 selected tests| View ReportGeneric LLM | Qwen3.8-27BConflicting evidence | HumanEval100 problems | 98 %Reported pass@1 · n = 100 | — | 42.1 tok/sGeneration · median · per request | 1 × W7800Single GPU · 48 GB VRAM/GPUConfiguration
|
|---|---|---|---|---|---|---|
| View Reportllm-bench.io | Qwen3.8-27B | llm-bench.io QualityFour sequential scenarios: agent workflow, code generation, role play, research | 85.56 / 100LLM judge · Mean | — | 61.1 tok/sGeneration · mean · per request | GPU count unknownUnknown · 24 GB VRAM/GPUConfiguration
|
| View Reportllm-bench.io | Qwen3.8-27B | llm-bench.io QualityFour sequential scenarios: agent workflow, code generation, role play, research | 72.53 / 100LLM judge · Mean | — | 18.58 tok/sGeneration · mean · per request | GPU count unknownUnknown · 16 GB VRAM/GPUConfiguration
|
| View Reportllm-bench.io | Qwen3.8-27B | llm-bench.io QualityFour sequential scenarios: agent workflow, code generation, role play, research | 81.27 / 100LLM judge · Mean | 171.11 tok/sPrefill · mean · per request | 5.78 tok/sGeneration · mean · per request | GPU count unknownUnknown · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Tool-use ladderDepth rung 1 of7; 1 profile(s) · schema-v2 / repeated-v116,784 input · 1 concurrent request | 100 %Checks passed · 1 / 1 · Tool-use checks | — | — | 2 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | WikiText128K configured context | 9.7297 perplexityWord perplexity · Perplexity · Lower is better | — | — | 1 × W7800Single GPU · 48 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | WikiText16 chunks of512 tokens1 concurrent request | 7.2 perplexityWord perplexity · Perplexity · Lower is better | — | — | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | LongBench v225 short/medium problems, seed42; mixed easy/hard1 concurrent request | 56 %Score · n = 100 | — | — | 1 × RX 9070 XTSingle GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportCustom evaluation | Qwen3.8-27B | GSM8KFirst 500 test questions; two few-shot seeds256K configured context | 97.5 %Accuracy · n = 100 · Flexible exact match · Mean | — | — | 2 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27BCorrectness concern | Tool schema30 forced structured-output requests | 30 passesScore · 30 / 30 · Schema-validation passes | — | — | 2 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportCustom evaluation | Qwen3.8-27B | HumanEval40 problems;every fourthHumanEval index · HumanEvalPlus; revision not reported1 concurrent request | 90 %pass@1 · 36 / 40 · Executable tests | — | — | 1 × Ryzen AI Max+ 395Single GPU · Unified memory · 128 GB RAMConfiguration
|
| View ReportCoding session | Qwen3.8-27B | Token speed | 14.2 tok/sThroughput · per request | 2,061.11 tok/sPrefill · per request | 24.7 tok/sGeneration · per request | 2 × RX 9070 XTLayer split · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | 1,168.6 tok/sPrefill · median · per request | 37.81 tok/sGeneration · median · per request | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportVS Code agent log | Qwen3.8-27B | Token speedMoorhuhn | — | 429.2 tok/sPrefill · per request | 34.14 tok/sGeneration · per request | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportVS Code agent log | Qwen3.8-27B | Token speedClair Obscure | — | 180.95 tok/sPrefill · per request | 15.55 tok/sGeneration · per request | 1 × R9700Single GPU · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | 898.58 tok/sPrefill · median · per request | 30.82 tok/sGeneration · median · per request | 1 × W7900Single GPU · 48 GB VRAM/GPUConfiguration
|
| View ReportBetterBench | Qwen3.8-27B | Token speed0.2.3 · chat | — | 1,000.5 tok/sPrefill · median · per request | 67.6 tok/sGeneration · median · per request | 2 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | 868.62 tok/sPrefill · median · per request | 27.89 tok/sGeneration · median · per request | 1 × W7900Single GPU · 48 GB VRAM/GPUConfiguration
|
| View ReportLocalMaxxing | Qwen3.8-27B | Token speed | — | 51.29 tok/sPrefill · per request | 154.18 tok/sGeneration · per request | 1 × RX 7900 XTXSingle GPU · 24 GB VRAM/GPUConfiguration
|
| View ReportLocalMaxxing | Qwen3.8-27B | Token speed | — | 500 tok/sPrefill · unknown | 42 tok/sGeneration · unknown | 1 × RX 7800 XTSingle GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27B | Token speed | — | 1,016.56 tok/sPrefill · per request | 27.55 tok/sGeneration · per request | Mixed: RX7900XTX24GB + RX7700XT12GB · 2 GPUsPipeline parallelConfiguration
|
| View ReportGeneric LLM | Qwen3.8-27BConflicting evidence | Token speed | — | 1,458 tok/sPrefill · per request | 29.7 tok/sGeneration · per request | 2 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
| View ReportCoding session | Qwen3.8-27B | Token speed | — | 246.8 tok/sPrefill · per request | 34.08 tok/sGeneration · per request | 1 × RX 7900 GRESingle GPU · 16 GB VRAM/GPUConfiguration
|
| View ReportLocalMaxxing | Qwen3.8-27B | Token speed | — | 1,470.64 tok/sPrefill · median · unknown | 51.79 tok/sGeneration · median · unknown | 2 × R9700Tensor parallel · 32 GB VRAM/GPUConfiguration
|
Quantizations and software
Quantizations: AD-IQ3_S_IQ3_XXS, BF16, Custom mixed 2.72 BPW (pareto-bf16), FP4, FP8, FP8 (native block 128x128), FP8 block (official; mixed BF16 modules), FP8 E4M3 block128, INT4 GPTQ v3, INT4 W4A16, INT5 ParoQuant W5A8, IQ3_S, IQ3_XXS, IU4, MQ4, MQ4V2, MXFP4, MXFP4 AWQ W4A8, MXFP4 ParoQuant W4A8, MXFP4 W4A8, MXFP4-hybrid-attnQ6, MXFP4-requant (dense), MXFP4-requant (UD), MXFP6 E2M3 W6A8, Native block FP8, NVFP4 checkpoint converted to MXFP4 at load, NVFP4-native, Q3_K_XL, Q4_0, Q4_0_ROCMFP4_STRIX, Q4_0_ROCMFP4_STRIX_LEAN, Q4_K_M, Q4_K_XL, Q6_K, Q8, Q8_0, Q8_K_XL, Quark AWQ MXFP4, UD-IQ4_XS, UD-Q3_K_XL, UD-Q4_K_M, UD-Q4_K_XL, UD-Q5_K_XL, UD-Q6_K, UD-Q6_K_M, UD-Q6_K_XL, UD-Q8_K_XL, Unsloth-Dynamic-IQ4_XS, W4A16 AutoRound GPTQ, W4A16-AutoRound-GPTQ
Engines: Gufo, hipEngine, hipfire, llama-bench, llama.cpp, llama.cpp / llama-server, llama.cpp RDNA lab fork, LM Studio, Loculus Core, lucebox, Ollama, Radiance, SGLang, Unsloth Studio, vllm, vLLM, vLLM / Radiance, vLLM + r9700-stack, vLLM Radiance, vLLM Radiance MXFP4, vLLM_for_AMD
This is a selection across sources, not a ranking or a matched experiment. Reading benchmark reports
GPU benchmark pages: W7900 · W7800 · R9700 · RX 9070 XT · RX 9070 · RX 9060 XT · RX 7900 XTX · RX 7900 XT · RX 7900 GRE · RX 7800 XT · RX 7600 XT · Ryzen AI Max+ 395