{"name":"Nemotron-3-Nano-4B","arena":{"id":"nemotron-3-nano-4b-fp8","name":"Nemotron-3-Nano 4B","vram":4,"params":4,"vendor":"nvidia","quality":{"coding":{"bench":"LiveCodeBench","value":51.8},"science":{"bench":"GPQA-Diamond","value":51.3},"knowledge":{"bench":"MMLU-Pro","value":18.1}},"contextK":256,"baseModelId":"nemotron-3-nano-4b"},"notes":"Model-ID: nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8. Container: vllm/vllm-openai:v0.26.0-aarch64-cu129-ubuntu2404. llama-benchy 0.4.0. Bronrun: 2026-08-08T09:26:41+02:00. KV-cache fp8_e4m3; max_model_len 49152; gpu_memory_utilization 0,9. Prefix caching uit; async scheduling aan. Extra serverflags: --trust-remote-code. 11 van de 11 tests hebben een meting die volgens de huidige Arena-regels meetelt. Bij deze historische run is geen sanity-check vastgelegd. Dat bewijs ontbreekt, ook wanneer de snelheidsmeting volgens de huidige regels meetelt.","order":220,"engine":"vLLM v0.26.0","verdict":"Gemeten op 2026-08-08 met vLLM v0.26.0. In de chattest met tien gelijktijdige verzoeken haalt dit profiel 34,89 tokens/s per gebruiker, met gemiddeld 0,93 seconden tot het eerste token. Bij 25k context stijgt die gemiddelde wachttijd naar 12,4 seconden. Bij deze historische run is geen sanity-check vastgelegd. Dat bewijs ontbreekt, ook wanneer de snelheidsmeting volgens de huidige regels meetelt.","hardware":"DGX Spark, NVIDIA GB10, 128 GB unified memory","modelUrl":"https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16","provider":"NVIDIA","precision":"FP8","parameters":"4B","provenance":{"modelId":"nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8","runPath":"results/nemotron-3/nemotron-3-nano-4b/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":"not_recorded"},"sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","servedName":"nemotron-3-nano-4b-fp8","generatedAt":"2026-08-08T09:26:41+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":48,"benchmarkResults":[{"id":"nemotron-3-nano-4b-fp8-chat-1k","date":"2026-08-08","runs":3,"metrics":{"ttfrSec":{"mean":0.931,"stddev":0.252},"decodeTokensPerSecondTotal":{"mean":308.16,"stddev":9.01},"prefillTokensPerSecondTotal":{"mean":7495.97,"stddev":71},"decodeTokensPerSecondPerUser":{"mean":34.89,"stddev":0.91}},"sourceRun":"Test 01","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","workloadId":"chat-1k","sourceLabel":"nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"01-chat","prefixCaching":false},{"id":"nemotron-3-nano-4b-fp8-rag-8k","date":"2026-08-08","runs":3,"metrics":{"ttfrSec":{"mean":3.853,"stddev":1.967},"decodeTokensPerSecondTotal":{"mean":225.13,"stddev":2.09},"prefillTokensPerSecondTotal":{"mean":10546.43,"stddev":199.19},"decodeTokensPerSecondPerUser":{"mean":27.64,"stddev":2.6}},"sourceRun":"Test 02","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","workloadId":"rag-8k","sourceLabel":"nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"02-rag-8k","prefixCaching":false},{"id":"nemotron-3-nano-4b-fp8-long-output-agents","date":"2026-08-08","runs":3,"metrics":{"ttfrSec":{"mean":0.331,"stddev":0.108},"decodeTokensPerSecondTotal":{"mean":197.08,"stddev":21.22},"prefillTokensPerSecondTotal":{"mean":6882.06,"stddev":1584.03},"decodeTokensPerSecondPerUser":{"mean":37.13,"stddev":1.9}},"sourceRun":"Test 03","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","workloadId":"long-output-agents","sourceLabel":"nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"03-long-output","prefixCaching":false},{"id":"nemotron-3-nano-4b-fp8-multi-turn-office","date":"2026-08-08","runs":3,"metrics":{"ttfrSec":{"mean":1.246,"stddev":0.571},"decodeTokensPerSecondTotal":{"mean":300.32,"stddev":3.53},"prefillTokensPerSecondTotal":{"mean":8845.2,"stddev":253.63},"decodeTokensPerSecondPerUser":{"mean":32.63,"stddev":1.05}},"sourceRun":"Test 04","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","workloadId":"multi-turn-office","sourceLabel":"nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"04-multi-turn","prefixCaching":false},{"id":"nemotron-3-nano-4b-fp8-large-context-25k","date":"2026-08-08","runs":3,"metrics":{"ttfrSec":{"mean":12.397,"stddev":6.513},"decodeTokensPerSecondTotal":{"mean":87.36,"stddev":0.9},"prefillTokensPerSecondTotal":{"mean":9914.35,"stddev":44.61},"decodeTokensPerSecondPerUser":{"mean":16.4,"stddev":6.03}},"sourceRun":"Test 05","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","workloadId":"large-context-25k","sourceLabel":"nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"05-big-context","prefixCaching":false},{"id":"nemotron-3-nano-4b-fp8-concurrency-stress-25k","date":"2026-08-08","runs":3,"metrics":{"ttfrSec":{"mean":24.235,"stddev":13.619},"decodeTokensPerSecondTotal":{"mean":90.74,"stddev":2.02},"prefillTokensPerSecondTotal":{"mean":9695.82,"stddev":275.59},"decodeTokensPerSecondPerUser":{"mean":10.09,"stddev":4.98}},"sourceRun":"Test 06","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","workloadId":"concurrency-stress-25k","sourceLabel":"nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 fp8 op de DGX Spark","sourceStatus":"complete","sourceTestId":"06-concurrency-stress","prefixCaching":false}],"openLoopResults":[{"id":"nemotron-3-nano-4b-fp8-office-random-4k","date":"2026-08-08","metrics":{"e2eMs":{"p50":15275.75,"p90":26808.1,"p95":29407.12,"p99":34158.79,"mean":15700.08},"tpotMs":{"p50":30.07,"p90":37.34,"p95":40.16,"p99":44.77,"mean":30.71},"ttftMs":{"p50":732.73,"p90":1308.53,"p95":1559.29,"p99":1736.3,"mean":810.13}},"sourceRun":"Test 07","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","scenarioId":"office-random-4k","achievedRps":0.292,"sourceStatus":"complete","sourceTestId":"07-office-baseline","configuredRps":0.3,"prefixCaching":false,"totalRequests":200,"successfulRequests":199,"totalTokenThroughput":1330.22,"peakConcurrentRequests":13},{"id":"nemotron-3-nano-4b-fp8-sharegpt-replay","date":"2026-08-08","metrics":{"e2eMs":{"p50":3194.98,"p90":11829.57,"p95":15055.98,"p99":19393.28,"mean":5006.8},"tpotMs":{"p50":23.33,"p90":24.87,"p95":25.55,"p99":27.5,"mean":23.49},"ttftMs":{"p50":107.44,"p90":403.15,"p95":477.89,"p99":675.55,"mean":157.31}},"sourceRun":"Test 08","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","scenarioId":"sharegpt-replay","achievedRps":0.3,"sourceStatus":"complete","sourceTestId":"08-sharegpt","configuredRps":0.3,"prefixCaching":false,"totalRequests":250,"successfulRequests":250,"totalTokenThroughput":132.26,"peakConcurrentRequests":8},{"id":"nemotron-3-nano-4b-fp8-reasoning","date":"2026-08-08","metrics":{"e2eMs":{"p50":186021.62,"p90":284677.37,"p95":317633.87,"p99":328936.22,"mean":183406.1},"tpotMs":{"p50":45.86,"p90":50.07,"p95":50.28,"p99":52.15,"mean":43.97},"ttftMs":{"p50":467.19,"p90":654.46,"p95":679.78,"p99":980.95,"mean":440.49}},"sourceRun":"Test 09","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","scenarioId":"reasoning","achievedRps":0.101,"sourceStatus":"complete","sourceTestId":"09-reasoning","configuredRps":0.2,"prefixCaching":false,"totalRequests":50,"successfulRequests":50,"totalTokenThroughput":535.91,"peakConcurrentRequests":35},{"id":"nemotron-3-nano-4b-fp8-monday-peak-random-4k","date":"2026-08-08","metrics":{"e2eMs":{"p50":33184.18,"p90":60985.37,"p95":63045.29,"p99":65736.93,"mean":33927.19},"tpotMs":{"p50":67.95,"p90":73.02,"p95":77.62,"p99":81.54,"mean":66.71},"ttftMs":{"p50":774.09,"p90":1334.11,"p95":1726.35,"p99":2167.16,"mean":840.95}},"sourceRun":"Test 10","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","scenarioId":"monday-peak-random-4k","achievedRps":0.702,"sourceStatus":"complete","sourceTestId":"10-monday-peak","configuredRps":1.5,"prefixCaching":false,"totalRequests":300,"successfulRequests":300,"totalTokenThroughput":3166.56,"peakConcurrentRequests":28}],"rateSweep":{"date":"2026-08-08","capacity":{"slo2000":{"outputTps":308.6,"ttftP95Ms":1574.3,"achievedRps":0.632,"configuredRps":0.7,"peakConcurrent":28},"slo5000":{"outputTps":427.4,"ttftP95Ms":2350.4,"achievedRps":0.869,"configuredRps":1,"peakConcurrent":75},"slo10000":{"outputTps":427.4,"ttftP95Ms":2350.4,"achievedRps":0.869,"configuredRps":1,"peakConcurrent":75}},"sourceRun":"Test 11","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/nemotron-3/nemotron-3-nano-4b/fp8/","abortedAtRps":null,"queueKneeRps":null,"sourceStatus":"complete","sourceTestId":"11-rate-sweep","totalRequests":850,"successfulRequests":850}}