{"id":"run_E0qv9whjqXf4YM7t","owner":"anonymous-EVjB2S","publishedAt":"2026-08-29T12:46:45.841Z","manifest":{"schema_version":"1.0","kind":"run","client_run_id":"rps-58623b0408c862bc37557f20cbdc945a9e42b0d5","name":"unsloth/Qwen3.8-27B-NVFP4 | baseline | 256->1024 | C32","notes":"Aggregate-only synthetic online serving measurement. It reports serving performance and does not evaluate model quality.","visibility":"public","license":"CC-BY-4.0","outcome":"completed","subject":{"type":"model","source":"huggingface","adapters":[],"revision":"7d6f8d4d72f56b92b3cdbf22f156b90e1bab0108","identifier":"unsloth/Qwen3.8-27B-NVFP4","modalities":["text"],"architecture":"Qwen3_5ForConditionalGeneration","quantization":{"format":"NVFP4 mixed precision","recipe":"Unsloth Dynamic V3.0 preview","checkpoint":"compressed-tensors"},"modifications":["Multimodal inputs disabled at serving time."],"revision_kind":"commit","parameter_count_total":27000000000,"parameter_count_active":27000000000},"environment":{"os":"Linux-6.8.0-136-generic-x86_64-with-glibc2.39","cpu":{"model":"x86_64"},"kernel":"6.8.0-136-generic","topology":"One server engine on one GPU.","accelerators":[{"count":1,"model":"NVIDIA GeForce RTX 5090","driver":"580.173.02","vendor":"NVIDIA","architecture":"Blackwell SM120","interconnect":"single device","memory_bytes_per_device":34190917632}],"driver_versions":{"nvidia":"580.173.02"},"host_architecture":"x86_64","system_memory_bytes":269585743872},"runtime":{"engine":"vllm","version":"0.11.2.dev280+gilded.gnosis.v20.vllm4d006a4.b12xcd3ce19.fi1ac6942.cu132.20260810.r34","extensions":{"runpile.server_variant":"baseline"},"parameters":{"server":{"dtype":"bfloat16","max_num_seqs":32,"enforce_eager":true,"max_model_len":9216,"kv_cache_dtype":"fp8","prefix_caching":false,"chunked_prefill":true,"expert_parallel":false,"async_scheduling":false,"gpu_memory_utilization":0.92,"max_num_batched_tokens":16384}},"parallelism":{"data":1,"expert":1,"tensor":1,"pipeline":1,"replicas":1},"launch_command":"CUDA_VISIBLE_DEVICES=0 vllm serve unsloth/Qwen3.8-27B-NVFP4 --revision 7d6f8d4d72f56b92b3cdbf22f156b90e1bab0108 --port 9000 --host 127.0.0.1 --dtype bfloat16 --gpu-memory-utilization 0.92 --max-model-len 9216 --max-num-batched-tokens 16384 --max-num-seqs 32 --tensor-parallel-size 1 --pipeline-parallel-size 1 --kv-cache-dtype fp8 --distributed-executor-backend mp --no-enable-prefix-caching --enable-chunked-prefill --no-async-scheduling --enforce-eager --served-model-name unsloth/Qwen3.8-27B-NVFP4 --disable-uvicorn-access-log --language-model-only","build_identifier":"voipmonitor/vllm@sha256:820181fbbc975cd5291c411cda9771d58fecee1636d916f508f47230df20592b","safe_environment":{"HF_HUB_OFFLINE":"1","CUDA_VISIBLE_DEVICES":"0","TOKENIZERS_PARALLELISM":"false"}},"workload":{"mode":"online","task":"text_generation","client":{"name":"vllm bench serve","backend":"vllm","version":"0.11.2.dev280+gilded.gnosis.v20.vllm4d006a4.b12xcd3ce19.fi1ac6942.cu132.20260810.r34","placement":"same-host"},"dataset":{"sha256":"9e1873a44a42e2cc60e0aef3613fb134f3786e4c50b64cd016ccf7ee9e65c0cb","generator":"runpile-sweeps-tokenizer-verified-jsonl","synthetic":true,"generator_revision":"d93886d7ee93eb0634b5f426ff480ce0a5798ad447d080e9c42a0d330a8bd055","shared_prefix_tokens":0},"sampling":{"mode":"greedy","temperature":0},"stopping":{"ignore_eos":true,"max_output_tokens":1024},"streaming":true,"tokenizer":{"revision":"7d6f8d4d72f56b92b3cdbf22f156b90e1bab0108","identifier":"unsloth/Qwen3.8-27B-NVFP4"},"extensions":{"runpile.profile_id":"serve-fixed-256-1k-v1"},"concurrency":32,"random_seed":0,"input_tokens":{"kind":"fixed","value":256},"endpoint_type":"openai-compatible-completions","output_tokens":{"kind":"fixed","value":1024},"request_count":128,"arrival_process":"closed_loop","prefix_behavior":{"cache_state":"warm-host","cache_enabled":false,"shared_prefix_tokens":0},"warmup_request_count":8,"request_rate_per_second":null},"measurement":{"notes":"Aggregate-only output.","warmup":"8 unmeasured warmups per invocation after one readiness request.","harness":"vllm bench serve","version":"0.11.2.dev280+gilded.gnosis.v20.vllm4d006a4.b12xcd3ce19.fi1ac6942.cu132.20260810.r34","extensions":{"runpile.declared_warmup_count":8,"runpile.max_gpu_temperature_c":54,"runpile.preemption_log_mentions":0,"runpile.readiness_request_count":1,"runpile.controller_source_sha256":"d93886d7ee93eb0634b5f426ff480ce0a5798ad447d080e9c42a0d330a8bd055","runpile.independent_repetition_index":1},"repetitions":1,"timing_scope":"client-observed streaming request latency","benchmark_command":"vllm bench serve --backend vllm --base-url http://127.0.0.1:9000 --endpoint /v1/completions --model unsloth/Qwen3.8-27B-NVFP4 --tokenizer unsloth/Qwen3.8-27B-NVFP4 --dataset-name custom --dataset-path <EXACT_SYNTHETIC_DATASET> --custom-output-len 1024 --skip-chat-template --disable-shuffle --num-prompts 128 --max-concurrency 32 --request-rate inf --seed 0 --ignore-eos --temperature 0 --percentile-metrics ttft,tpot,itl,e2el --metric-percentiles 50,90,95,99 --disable-tqdm --save-result --result-dir <local-path> --result-filename qwen38-27b-nvfp4-gg-r34-tp1__baseline__serve-fixed-256-1k-v1__c32__r1__attempt1.json --num-warmups 8 --metadata campaign=runpile-rtx5090-qwen38-exl3-pilot-20260828 config_id=qwen38-27b-nvfp4-gg-r34-tp1 server_variant=baseline workload_key=serve-fixed-256-1k-v1 repetition=1 aggregate_only=true","retry_failure_treatment":"Failed requests counted; no client retries.","client_overhead_included":true},"results":{"latency_ms":{"end_to_end":{"p50":39627.883735112846,"p90":78724.38268307596,"p95":78968.34144494496,"p99":79011.74961589277,"mean":45580.63611235775,"stddev":14223.935309659362,"sample_count":128,"measurement_status":"measured"},"inter_token":{"p50":37.93093212880194,"p90":38.32946438342333,"p95":38.497228105552495,"p99":38.9390378119424,"mean":38.08229122833664,"stddev":7.2369170627985735,"sample_count":128,"measurement_status":"measured"},"time_to_first_token":{"p50":709.8625309299678,"p90":39471.25625079498,"p95":39587.053017900325,"p99":39752.24923652131,"mean":6622.452305586194,"stddev":14137.782680196637,"sample_count":128,"measurement_status":"measured"},"time_per_output_token":{"p50":37.958043217995844,"p90":38.39465613569791,"p95":38.56927828229636,"p99":38.597381557397114,"mean":38.08229111121364,"stddev":0.2531707382807226,"sample_count":128,"measurement_status":"measured"}},"throughput":{"requests_per_second":{"value":0.6503127669587239,"sample_count":128,"measurement_status":"measured"},"total_tokens_per_second":{"value":832.4003417071666,"sample_count":128,"measurement_status":"measured"},"output_tokens_per_second":{"value":665.9202733657332,"sample_count":128,"measurement_status":"measured"}},"failed_requests":0,"duration_seconds":196.828367061913,"total_input_tokens":32768,"successful_requests":128,"total_output_tokens":131072},"artifacts":[{"kind":"vllm_bench_serve_json","sha256":"750cd974c0b600066f606668bcc9106317e2c5a5456d26f0f257fca55464456d","license":"CC-BY-4.0","filename":"qwen38-27b-nvfp4-gg-r34-tp1__baseline__serve-fixed-256-1k-v1__c32__r1.json","media_type":"application/json","size_bytes":1912,"visibility":"private","client_artifact_id":"raw-vllm-serve","raw_requests_included":false}],"provenance":{"tool":"runpile HTTP importer","source":"import","tool_version":"1.0"},"warnings":["Review imported facts against the local model, hardware, launch command, and benchmark command before publication."]},"indexingState":"indexed","resultKey":"res_xXGPXNIAQqKJfwSlIdrqzuLY","eligibilityReasons":[],"artifactBacked":true,"downloads":[]}