{"name":"Qwen-3.5 4B","arena":{"id":"qwen-3-5-4b-bf16","name":"Qwen-3.5 4B","vram":8,"params":4,"vendor":"alibaba","quality":{"coding":{"bench":"LiveCodeBench v6","value":55.8},"science":{"bench":"GPQA-Diamond","value":76.2},"knowledge":{"bench":"MMLU-Pro","value":79.1}},"contextK":256},"notes":"Model-ID: Qwen/Qwen3.5-4B. Container: vllm/vllm-openai:v0.26.0-aarch64-cu129-ubuntu2404. llama-benchy 0.4.0. Bronrun: 2026-08-07T16:20:14+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. 11 van de 11 tests hebben een meting die volgens de huidige Arena-regels meetelt. Er is output voor de sanity-check opgeslagen, maar die is nog niet beoordeeld. De huidige telling betekent dus niet dat die controle is geslaagd.","order":300,"engine":"vLLM v0.26.0","verdict":"Gemeten op 2026-08-07 met vLLM v0.26.0. In de chattest met tien gelijktijdige verzoeken haalt dit profiel 22,02 tokens/s per gebruiker, met gemiddeld 0,99 seconden tot het eerste token. Bij 25k context stijgt die gemiddelde wachttijd naar 23,05 seconden. Er is output voor de sanity-check opgeslagen, maar die is nog niet beoordeeld. De huidige telling betekent dus niet dat die controle is geslaagd.","hardware":"DGX Spark, NVIDIA GB10, 128 GB unified memory","modelUrl":"https://huggingface.co/Qwen/Qwen3.5-4B","provider":"Qwen","precision":"BF16","parameters":"4B","provenance":{"modelId":"Qwen/Qwen3.5-4B","runPath":"results/qwen-3.5/qwen-3.5-4b/bf16","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":"captured_unreviewed"},"sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","servedName":"qwen3.5-4b-bf16","generatedAt":"2026-08-07T16:20:14+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":"Hybrid","measuredContextK":128,"benchmarkResults":[{"id":"qwen-3-5-4b-bf16-chat-1k","date":"2026-08-07","runs":3,"metrics":{"ttfrSec":{"mean":0.987,"stddev":0.441},"decodeTokensPerSecondTotal":{"mean":216.04,"stddev":2.63},"prefillTokensPerSecondTotal":{"mean":6024.65,"stddev":25.94},"decodeTokensPerSecondPerUser":{"mean":22.02,"stddev":0.33}},"sourceRun":"Test 01","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","workloadId":"chat-1k","sourceLabel":"Qwen/Qwen3.5-4B bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"01-chat","prefixCaching":false},{"id":"qwen-3-5-4b-bf16-rag-8k","date":"2026-08-07","runs":3,"metrics":{"ttfrSec":{"mean":7.145,"stddev":3.558},"decodeTokensPerSecondTotal":{"mean":142.18,"stddev":0.49},"prefillTokensPerSecondTotal":{"mean":5895.33,"stddev":34.81},"decodeTokensPerSecondPerUser":{"mean":17.5,"stddev":1.85}},"sourceRun":"Test 02","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","workloadId":"rag-8k","sourceLabel":"Qwen/Qwen3.5-4B bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"02-rag-8k","prefixCaching":false},{"id":"qwen-3-5-4b-bf16-long-output-agents","date":"2026-08-07","runs":3,"metrics":{"ttfrSec":{"mean":0.389,"stddev":0.098},"decodeTokensPerSecondTotal":{"mean":219.32,"stddev":1.43},"prefillTokensPerSecondTotal":{"mean":5308.55,"stddev":330.47},"decodeTokensPerSecondPerUser":{"mean":22.1,"stddev":0.05}},"sourceRun":"Test 03","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","workloadId":"long-output-agents","sourceLabel":"Qwen/Qwen3.5-4B bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"03-long-output","prefixCaching":false},{"id":"qwen-3-5-4b-bf16-multi-turn-office","date":"2026-08-07","runs":3,"metrics":{"ttfrSec":{"mean":1.732,"stddev":0.879},"decodeTokensPerSecondTotal":{"mean":198.88,"stddev":1.74},"prefillTokensPerSecondTotal":{"mean":6290.96,"stddev":23.54},"decodeTokensPerSecondPerUser":{"mean":21.22,"stddev":0.69}},"sourceRun":"Test 04","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","workloadId":"multi-turn-office","sourceLabel":"Qwen/Qwen3.5-4B bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"04-multi-turn","prefixCaching":false},{"id":"qwen-3-5-4b-bf16-large-context-25k","date":"2026-08-07","runs":3,"metrics":{"ttfrSec":{"mean":23.047,"stddev":12.022},"decodeTokensPerSecondTotal":{"mean":49.14,"stddev":0.22},"prefillTokensPerSecondTotal":{"mean":5389.14,"stddev":30.76},"decodeTokensPerSecondPerUser":{"mean":9.5,"stddev":3.82}},"sourceRun":"Test 05","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","workloadId":"large-context-25k","sourceLabel":"Qwen/Qwen3.5-4B bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"05-big-context","prefixCaching":false},{"id":"qwen-3-5-4b-bf16-concurrency-stress-25k","date":"2026-08-07","runs":3,"metrics":{"ttfrSec":{"mean":44.648,"stddev":25.059},"decodeTokensPerSecondTotal":{"mean":51.26,"stddev":0.03},"prefillTokensPerSecondTotal":{"mean":5239.06,"stddev":1.34},"decodeTokensPerSecondPerUser":{"mean":5.94,"stddev":3.19}},"sourceRun":"Test 06","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","workloadId":"concurrency-stress-25k","sourceLabel":"Qwen/Qwen3.5-4B bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"06-concurrency-stress","prefixCaching":false}],"openLoopResults":[{"id":"qwen-3-5-4b-bf16-office-random-4k","date":"2026-08-07","metrics":{"e2eMs":{"p50":27476.73,"p90":47636.79,"p95":52356.34,"p99":55766.89,"mean":27577.28},"tpotMs":{"p50":54.67,"p90":63.9,"p95":67.76,"p99":73.1,"mean":54.97},"ttftMs":{"p50":961.29,"p90":1744.18,"p95":2163.37,"p99":2685.26,"mean":1026.65}},"sourceRun":"Test 07","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","scenarioId":"office-random-4k","achievedRps":0.287,"sourceStatus":"complete","sourceTestId":"07-office-baseline","configuredRps":0.3,"prefixCaching":false,"totalRequests":200,"successfulRequests":200,"totalTokenThroughput":1308.91,"peakConcurrentRequests":16},{"id":"qwen-3-5-4b-bf16-sharegpt-replay","date":"2026-08-07","metrics":{"e2eMs":{"p50":5891.39,"p90":20757.12,"p95":27277.01,"p99":34892.54,"mean":8887.09},"tpotMs":{"p50":41.25,"p90":43.29,"p95":45.27,"p99":48.03,"mean":41.53},"ttftMs":{"p50":145.76,"p90":195.5,"p95":228.77,"p99":250.67,"mean":150.36}},"sourceRun":"Test 08","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","scenarioId":"sharegpt-replay","achievedRps":0.298,"sourceStatus":"complete","sourceTestId":"08-sharegpt","configuredRps":0.3,"prefixCaching":false,"totalRequests":250,"successfulRequests":250,"totalTokenThroughput":131.09,"peakConcurrentRequests":10},{"id":"qwen-3-5-4b-bf16-reasoning","date":"2026-08-07","metrics":{"e2eMs":{"p50":273083.83,"p90":419932.75,"p95":461617.87,"p99":474695.08,"mean":274285.66},"tpotMs":{"p50":66.51,"p90":72.48,"p95":73.7,"p99":76.04,"mean":65.72},"ttftMs":{"p50":317.71,"p90":437.82,"p95":460.07,"p99":508.78,"mean":319.36}},"sourceRun":"Test 09","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","scenarioId":"reasoning","achievedRps":0.077,"sourceStatus":"complete","sourceTestId":"09-reasoning","configuredRps":0.2,"prefixCaching":false,"totalRequests":50,"successfulRequests":50,"totalTokenThroughput":406.93,"peakConcurrentRequests":43},{"id":"qwen-3-5-4b-bf16-monday-peak-random-4k","date":"2026-08-07","metrics":{"e2eMs":{"p50":44785.48,"p90":81976.46,"p95":84943.45,"p99":89442.35,"mean":45898.68},"tpotMs":{"p50":91.01,"p90":98.42,"p95":102.32,"p99":125.66,"mean":90.54},"ttftMs":{"p50":980.28,"p90":1761.24,"p95":2362.68,"p99":5327.14,"mean":1137.01}},"sourceRun":"Test 10","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","scenarioId":"monday-peak-random-4k","achievedRps":0.519,"sourceStatus":"complete","sourceTestId":"10-monday-peak","configuredRps":1.5,"prefixCaching":false,"totalRequests":300,"successfulRequests":300,"totalTokenThroughput":2340.29,"peakConcurrentRequests":28}],"rateSweep":{"date":"2026-08-07","capacity":{"slo2000":{"outputTps":47.1,"ttftP95Ms":1764.7,"achievedRps":0.096,"configuredRps":0.1,"peakConcurrent":8},"slo5000":{"outputTps":283.3,"ttftP95Ms":2967.2,"achievedRps":0.58,"configuredRps":0.7,"peakConcurrent":57},"slo10000":{"outputTps":362.2,"ttftP95Ms":6610.7,"achievedRps":0.736,"configuredRps":1,"peakConcurrent":150}},"sourceRun":"Test 11","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/qwen-3.5/qwen-3.5-4b/bf16/","abortedAtRps":null,"queueKneeRps":null,"sourceStatus":"complete","sourceTestId":"11-rate-sweep","totalRequests":850,"successfulRequests":850}}