{"name":"Qwen-3.8 27B","arena":{"id":"qwen-3-8-27b-fp8","name":"Qwen-3.8 27B","vram":30,"params":27,"vendor":"alibaba","quality":{"coding":{"bench":"LiveCodeBench v6","value":90.3},"science":{"bench":"GPQA-Diamond","value":89.2},"knowledge":null},"contextK":256,"baseModelId":"qwen-3-8-27b"},"notes":"Model-ID: Qwen/Qwen3.8-27B-FP8. Container: vllm/vllm-openai:v0.26.0-aarch64-cu129-ubuntu2404. llama-benchy 0.4.0. Bronrun: 2026-08-15T22:23:08+02:00. KV-cache fp8; max_model_len 131072; gpu_memory_utilization 0,9. Prefix caching uit; async scheduling aan. Extra serverflags: --language-model-only --reasoning-parser qwen3 --trust-remote-code --linear-backend marlin. Profielvariabelen: VLLM_TEST_FORCE_FP8_MARLIN=1. 11 van de 11 tests hebben een meting die volgens de huidige Arena-regels meetelt. De sanity-check is geslaagd. Dit is een controle op een kort gegenereerd antwoord, geen inhoudelijke evaluatie van de benchmarktaken.","order":380,"engine":"vLLM v0.26.0","verdict":"Gemeten op 2026-08-15 tot 2026-08-16 met vLLM v0.26.0. In de chattest met tien gelijktijdige verzoeken haalt dit profiel 7,06 tokens/s per gebruiker, met gemiddeld 10,86 seconden tot het eerste token. Bij 25k context stijgt die gemiddelde wachttijd naar 215,07 seconden. Deze runs meten snelheid en wachttijd, niet de kwaliteit van de antwoorden.","hardware":"DGX Spark, NVIDIA GB10, 128 GB unified memory","modelUrl":"https://huggingface.co/Qwen/Qwen3.8-27B-FP8","provider":"Qwen","precision":"FP8","parameters":"27B","provenance":{"modelId":"Qwen/Qwen3.8-27B-FP8","runPath":"results/qwen-3.8/qwen-3.8-27b/fp8","validity":{"tests":{"01-chat":"complete","02-rag-8k":"complete","08-sharegpt":"complete","09-reasoning":"complete","04-multi-turn":"complete","11-rate-sweep":"complete","03-long-output":"complete","05-big-context":"complete","10-monday-peak":"complete","07-office-baseline":"complete","06-concurrency-stress":"complete"},"sanity":"passed"},"sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","servedName":"qwen3.8-27b-fp8","generatedAt":"2026-08-15T22:23:08+02:00","runRevision":"a948cec23b054c62fef595bad3cdb284066922ac","vllmVersion":"0.26.0","suiteVersion":"2026-08","repositoryUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks","contractVersion":1,"contractRevision":"a948cec23b054c62fef595bad3cdb284066922ac","commandProvenance":"reconstructed_from_raw_artifacts","llamaBenchyVersion":"0.4.0"},"architecture":"Dense","measuredContextK":128,"benchmarkResults":[{"id":"qwen-3-8-27b-fp8-chat-1k","date":"2026-08-15","runs":3,"metrics":{"ttfrSec":{"mean":10.863,"stddev":4.022},"decodeTokensPerSecondTotal":{"mean":66.85,"stddev":0.04},"prefillTokensPerSecondTotal":{"mean":606.85,"stddev":0.66},"decodeTokensPerSecondPerUser":{"mean":7.06,"stddev":0.18}},"sourceRun":"Test 01","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","workloadId":"chat-1k","sourceLabel":"Qwen/Qwen3.8-27B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"01-chat","prefixCaching":false},{"id":"qwen-3-8-27b-fp8-rag-8k","date":"2026-08-15","runs":3,"metrics":{"ttfrSec":{"mean":70.145,"stddev":34.54},"decodeTokensPerSecondTotal":{"mean":28.13,"stddev":0.06},"prefillTokensPerSecondTotal":{"mean":608.66,"stddev":0.17},"decodeTokensPerSecondPerUser":{"mean":4.44,"stddev":1.27}},"sourceRun":"Test 02","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","workloadId":"rag-8k","sourceLabel":"Qwen/Qwen3.8-27B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"02-rag-8k","prefixCaching":false},{"id":"qwen-3-8-27b-fp8-long-output-agents","date":"2026-08-15","runs":3,"metrics":{"ttfrSec":{"mean":3.688,"stddev":0.542},"decodeTokensPerSecondTotal":{"mean":65.57,"stddev":1.95},"prefillTokensPerSecondTotal":{"mean":550.48,"stddev":6.67},"decodeTokensPerSecondPerUser":{"mean":7.29,"stddev":0.03}},"sourceRun":"Test 03","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","workloadId":"long-output-agents","sourceLabel":"Qwen/Qwen3.8-27B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"03-long-output","prefixCaching":false},{"id":"qwen-3-8-27b-fp8-multi-turn-office","date":"2026-08-15","runs":3,"metrics":{"ttfrSec":{"mean":19.286,"stddev":8.894},"decodeTokensPerSecondTotal":{"mean":52.63,"stddev":0.03},"prefillTokensPerSecondTotal":{"mean":612.17,"stddev":0.63},"decodeTokensPerSecondPerUser":{"mean":6.3,"stddev":0.65}},"sourceRun":"Test 04","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","workloadId":"multi-turn-office","sourceLabel":"Qwen/Qwen3.8-27B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"04-multi-turn","prefixCaching":false},{"id":"qwen-3-8-27b-fp8-large-context-25k","date":"2026-08-15","runs":3,"metrics":{"ttfrSec":{"mean":215.075,"stddev":110.153},"decodeTokensPerSecondTotal":{"mean":6.65,"stddev":0.02},"prefillTokensPerSecondTotal":{"mean":587.95,"stddev":0.44},"decodeTokensPerSecondPerUser":{"mean":1.96,"stddev":1.66}},"sourceRun":"Test 05","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","workloadId":"large-context-25k","sourceLabel":"Qwen/Qwen3.8-27B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"05-big-context","prefixCaching":false},{"id":"qwen-3-8-27b-fp8-concurrency-stress-25k","date":"2026-08-15","runs":3,"metrics":{"ttfrSec":{"mean":409.504,"stddev":223.999},"decodeTokensPerSecondTotal":{"mean":6.53,"stddev":0.02},"prefillTokensPerSecondTotal":{"mean":582.63,"stddev":0.11},"decodeTokensPerSecondPerUser":{"mean":1.14,"stddev":1.24}},"sourceRun":"Test 06","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","workloadId":"concurrency-stress-25k","sourceLabel":"Qwen/Qwen3.8-27B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"06-concurrency-stress","prefixCaching":false}],"openLoopResults":[{"id":"qwen-3-8-27b-fp8-office-random-4k","date":"2026-08-16","metrics":{"e2eMs":{"p50":1235227.37,"p90":1531025.35,"p95":1615442.23,"p99":1668491.21,"mean":1213973.83},"tpotMs":{"p50":1977.54,"p90":3511.85,"p95":3553.36,"p99":3613.44,"mean":2013.51},"ttftMs":{"p50":426054.6,"p90":718813.75,"p95":756550.09,"p99":809321.28,"mean":408512.28}},"sourceRun":"Test 07","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","scenarioId":"office-random-4k","achievedRps":0.112,"sourceStatus":"complete","sourceTestId":"07-office-baseline","configuredRps":0.3,"prefixCaching":false,"totalRequests":200,"successfulRequests":200,"totalTokenThroughput":515.43,"peakConcurrentRequests":196},{"id":"qwen-3-8-27b-fp8-sharegpt-replay","date":"2026-08-16","metrics":{"e2eMs":{"p50":20835.91,"p90":80195.12,"p95":100399.1,"p99":134804.83,"mean":33496.73},"tpotMs":{"p50":157.19,"p90":173.24,"p95":188.56,"p99":255.3,"mean":158.71},"ttftMs":{"p50":907.94,"p90":1762.6,"p95":2095.59,"p99":2632.11,"mean":1021.1}},"sourceRun":"Test 08","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","scenarioId":"sharegpt-replay","achievedRps":0.28,"sourceStatus":"complete","sourceTestId":"08-sharegpt","configuredRps":0.3,"prefixCaching":false,"totalRequests":250,"successfulRequests":250,"totalTokenThroughput":133.74,"peakConcurrentRequests":19},{"id":"qwen-3-8-27b-fp8-reasoning","date":"2026-08-16","metrics":{"e2eMs":{"p50":845969.19,"p90":1254962.89,"p95":1365229.04,"p99":1412105.91,"mean":858020.3},"tpotMs":{"p50":213.32,"p90":236.25,"p95":243.04,"p99":256.44,"mean":212.05},"ttftMs":{"p50":3069.14,"p90":5428.67,"p95":6486.25,"p99":8334.73,"mean":3416.02}},"sourceRun":"Test 09","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","scenarioId":"reasoning","achievedRps":0.033,"sourceStatus":"complete","sourceTestId":"09-reasoning","configuredRps":0.2,"prefixCaching":false,"totalRequests":50,"successfulRequests":50,"totalTokenThroughput":175.32,"peakConcurrentRequests":49},{"id":"qwen-3-8-27b-fp8-monday-peak-random-4k","date":"2026-08-16","metrics":{"e2eMs":{"p50":231434.61,"p90":440451.1,"p95":463910.3,"p99":534855.99,"mean":249234.66},"tpotMs":{"p50":477.65,"p90":561.45,"p95":620.28,"p99":831.84,"mean":480.35},"ttftMs":{"p50":8814.3,"p90":18253.81,"p95":55083.13,"p99":145158.18,"mean":14826.71}},"sourceRun":"Test 10","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","scenarioId":"monday-peak-random-4k","achievedRps":0.098,"sourceStatus":"complete","sourceTestId":"10-monday-peak","configuredRps":1.5,"prefixCaching":false,"totalRequests":300,"successfulRequests":300,"totalTokenThroughput":446.86,"peakConcurrentRequests":28}],"rateSweep":{"date":"2026-08-15","capacity":{"slo2000":null,"slo5000":null,"slo10000":null},"sourceRun":"Test 11","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.8/qwen-3.8-27b/fp8/","abortedAtRps":null,"queueKneeRps":0.1,"sourceStatus":"complete","sourceTestId":"11-rate-sweep","totalRequests":850,"successfulRequests":850}}