{"name":"Gemma-4 E4B","arena":{"id":"gemma-4-e4b-it-bf16","name":"Gemma-4 E4B","vram":9,"params":4.5,"vendor":"google","quality":{"coding":{"bench":"LiveCodeBench v6","value":52},"science":{"bench":"GPQA-Diamond","value":58.6},"knowledge":{"bench":"MMLU-Pro","value":69.4}},"contextK":128},"notes":"Model-ID: google/gemma-4-E4B-it. Container: vllm/vllm-openai:v0.26.0-aarch64-cu129-ubuntu2404. llama-benchy 0.4.0. Bronrun: 2026-08-17T03:07:46+02:00. KV-cache fp8; max_model_len 131072; gpu_memory_utilization 0,9. Prefix caching uit; async scheduling aan. Extra serverflags: --max-num-batched-tokens 8192. 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":80,"engine":"vLLM v0.26.0","verdict":"Gemeten op 2026-08-17 met vLLM v0.26.0. In de chattest met tien gelijktijdige verzoeken haalt dit profiel 21,47 tokens/s per gebruiker, met gemiddeld 1,05 seconden tot het eerste token. Bij 25k context stijgt die gemiddelde wachttijd naar 29,43 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/google/gemma-4-E4B-it","provider":"google","precision":"BF16","parameters":"4.5B effectief","provenance":{"modelId":"google/gemma-4-E4B-it","runPath":"results/gemma-4/gemma-4-e4b-it/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":"passed"},"sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","servedName":"gemma-4-e4b-it-bf16","generatedAt":"2026-08-17T03:07:46+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":"gemma-4-e4b-it-bf16-chat-1k","date":"2026-08-17","runs":3,"metrics":{"ttfrSec":{"mean":1.051,"stddev":0.399},"decodeTokensPerSecondTotal":{"mean":134.08,"stddev":24.97},"prefillTokensPerSecondTotal":{"mean":6766.94,"stddev":213.5},"decodeTokensPerSecondPerUser":{"mean":21.47,"stddev":1.99}},"sourceRun":"Test 01","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","workloadId":"chat-1k","sourceLabel":"google/gemma-4-E4B-it bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"01-chat","prefixCaching":false},{"id":"gemma-4-e4b-it-bf16-rag-8k","date":"2026-08-17","runs":3,"metrics":{"ttfrSec":{"mean":6.935,"stddev":3.542},"decodeTokensPerSecondTotal":{"mean":118.42,"stddev":7.12},"prefillTokensPerSecondTotal":{"mean":6271.18,"stddev":14.39},"decodeTokensPerSecondPerUser":{"mean":16.04,"stddev":4.2}},"sourceRun":"Test 02","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","workloadId":"rag-8k","sourceLabel":"google/gemma-4-E4B-it bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"02-rag-8k","prefixCaching":false},{"id":"gemma-4-e4b-it-bf16-long-output-agents","date":"2026-08-17","runs":3,"metrics":{"ttfrSec":{"mean":0.403,"stddev":0.003},"decodeTokensPerSecondTotal":{"mean":125.67,"stddev":4.94},"prefillTokensPerSecondTotal":{"mean":5897,"stddev":48.3},"decodeTokensPerSecondPerUser":{"mean":22.45,"stddev":0.21}},"sourceRun":"Test 03","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","workloadId":"long-output-agents","sourceLabel":"google/gemma-4-E4B-it bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"03-long-output","prefixCaching":false},{"id":"gemma-4-e4b-it-bf16-multi-turn-office","date":"2026-08-17","runs":3,"metrics":{"ttfrSec":{"mean":1.806,"stddev":0.728},"decodeTokensPerSecondTotal":{"mean":145.37,"stddev":15.16},"prefillTokensPerSecondTotal":{"mean":7190.94,"stddev":15.87},"decodeTokensPerSecondPerUser":{"mean":20.77,"stddev":1.25}},"sourceRun":"Test 04","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","workloadId":"multi-turn-office","sourceLabel":"google/gemma-4-E4B-it bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"04-multi-turn","prefixCaching":false},{"id":"gemma-4-e4b-it-bf16-large-context-25k","date":"2026-08-17","runs":3,"metrics":{"ttfrSec":{"mean":29.431,"stddev":15.009},"decodeTokensPerSecondTotal":{"mean":39.13,"stddev":1.45},"prefillTokensPerSecondTotal":{"mean":4275.82,"stddev":2.7},"decodeTokensPerSecondPerUser":{"mean":7.94,"stddev":4.01}},"sourceRun":"Test 05","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","workloadId":"large-context-25k","sourceLabel":"google/gemma-4-E4B-it bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"05-big-context","prefixCaching":false},{"id":"gemma-4-e4b-it-bf16-concurrency-stress-25k","date":"2026-08-17","runs":3,"metrics":{"ttfrSec":{"mean":55.76,"stddev":30.193},"decodeTokensPerSecondTotal":{"mean":41.47,"stddev":1.04},"prefillTokensPerSecondTotal":{"mean":4285.32,"stddev":4.69},"decodeTokensPerSecondPerUser":{"mean":4.74,"stddev":2.75}},"sourceRun":"Test 06","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","workloadId":"concurrency-stress-25k","sourceLabel":"google/gemma-4-E4B-it bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"06-concurrency-stress","prefixCaching":false}],"openLoopResults":[{"id":"gemma-4-e4b-it-bf16-office-random-4k","date":"2026-08-17","metrics":{"e2eMs":{"p50":28109.4,"p90":48800.28,"p95":54195.11,"p99":57904.89,"mean":28836.6},"tpotMs":{"p50":57.2,"p90":66.56,"p95":69.62,"p99":76.98,"mean":57.81},"ttftMs":{"p50":888.64,"p90":1713.88,"p95":2083.88,"p99":2698.97,"mean":972.17}},"sourceRun":"Test 07","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","scenarioId":"office-random-4k","achievedRps":0.286,"sourceStatus":"complete","sourceTestId":"07-office-baseline","configuredRps":0.3,"prefixCaching":false,"totalRequests":200,"successfulRequests":200,"totalTokenThroughput":1304.99,"peakConcurrentRequests":16},{"id":"gemma-4-e4b-it-bf16-sharegpt-replay","date":"2026-08-17","metrics":{"e2eMs":{"p50":5956.27,"p90":22163.59,"p95":30214.1,"p99":38088.38,"mean":9230.18},"tpotMs":{"p50":44.32,"p90":46.74,"p95":48.55,"p99":51.23,"mean":44.99},"ttftMs":{"p50":165.37,"p90":232.72,"p95":249.56,"p99":305.34,"mean":174.24}},"sourceRun":"Test 08","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","scenarioId":"sharegpt-replay","achievedRps":0.299,"sourceStatus":"complete","sourceTestId":"08-sharegpt","configuredRps":0.3,"prefixCaching":false,"totalRequests":250,"successfulRequests":250,"totalTokenThroughput":131.5,"peakConcurrentRequests":10},{"id":"gemma-4-e4b-it-bf16-reasoning","date":"2026-08-17","metrics":{"e2eMs":{"p50":250096.44,"p90":400208.23,"p95":438962.11,"p99":450848.54,"mean":258118.39},"tpotMs":{"p50":62.1,"p90":65.88,"p95":66.87,"p99":68.16,"mean":61.65},"ttftMs":{"p50":317.82,"p90":439.64,"p95":450.67,"p99":468.4,"mean":329.22}},"sourceRun":"Test 09","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","scenarioId":"reasoning","achievedRps":0.078,"sourceStatus":"complete","sourceTestId":"09-reasoning","configuredRps":0.2,"prefixCaching":false,"totalRequests":50,"successfulRequests":50,"totalTokenThroughput":414.4,"peakConcurrentRequests":43},{"id":"gemma-4-e4b-it-bf16-monday-peak-random-4k","date":"2026-08-17","metrics":{"e2eMs":{"p50":43695.74,"p90":79706.14,"p95":82590.52,"p99":86863.9,"mean":44743.19},"tpotMs":{"p50":88.61,"p90":96.44,"p95":101.02,"p99":129.7,"mean":88.74},"ttftMs":{"p50":808.65,"p90":1379.89,"p95":2021.93,"p99":5289.65,"mean":976.52}},"sourceRun":"Test 10","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","scenarioId":"monday-peak-random-4k","achievedRps":0.532,"sourceStatus":"complete","sourceTestId":"10-monday-peak","configuredRps":1.5,"prefixCaching":false,"totalRequests":300,"successfulRequests":300,"totalTokenThroughput":2395.19,"peakConcurrentRequests":28}],"rateSweep":{"date":"2026-08-17","capacity":{"slo2000":{"outputTps":90.2,"ttftP95Ms":1885.6,"achievedRps":0.184,"configuredRps":0.2,"peakConcurrent":15},"slo5000":{"outputTps":391.8,"ttftP95Ms":4062.7,"achievedRps":0.796,"configuredRps":1,"peakConcurrent":130},"slo10000":{"outputTps":391.8,"ttftP95Ms":4062.7,"achievedRps":0.796,"configuredRps":1,"peakConcurrent":130}},"sourceRun":"Test 11","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/gemma-4/gemma-4-e4b-it/bf16/","abortedAtRps":null,"queueKneeRps":null,"sourceStatus":"complete","sourceTestId":"11-rate-sweep","totalRequests":850,"successfulRequests":850}}