{"name":"Ministral-3 3B","arena":{"id":"ministral-3-3b-instruct-bf16","name":"Ministral-3 3B","vram":8,"params":3,"vendor":"mistral","quality":{"coding":{"bench":"LiveCodeBench","value":54.8},"science":{"bench":"GPQA-Diamond","value":53.4},"knowledge":{"bench":"MMLU","value":70.7}},"contextK":256},"notes":"Model-ID: mistralai/Ministral-3-3B-Instruct-2512. Container: vllm/vllm-openai:v0.26.0-aarch64-cu129-ubuntu2404. llama-benchy 0.4.0. Bronrun: 2026-08-09T00:19:16+02:00. KV-cache auto; max_model_len 131072; gpu_memory_utilization 0,9. Prefix caching uit; async scheduling aan. Extra serverflags: --tokenizer_mode mistral --config_format mistral --load_format mistral --max-num-batched-tokens 8192. 0 van de 11 tests hebben een meting die volgens de huidige Arena-regels meetelt. De sanity-check is mislukt. Deze run wordt daarom uitgesloten van de rangschikking; de ruwe cijfers blijven alleen ter inspectie zichtbaar. Afgebroken: 11-rate-sweep.","order":13,"engine":"vLLM v0.26.0","verdict":"De sanity-check is mislukt. Deze run wordt daarom uitgesloten van de rangschikking; de ruwe cijfers blijven alleen ter inspectie zichtbaar. Gemeten op 2026-08-08 tot 2026-08-09 met vLLM v0.26.0. De eerdere aanbeveling op basis van deze snelheidsresultaten is vervallen.","findings":{"broke":{"body":"Ook de test met 25k inputtokens is uitgesloten. De ruwe resultaten staan bij de test, maar vormen geen geldige basis voor een prestatieadvies.","title":"Wachttijd bij 25k context"},"costly":{"body":"De piekmeting is uitgesloten door de mislukte sanity-check. Een geslaagde HTTP-aanvraag is niet hetzelfde als een bruikbaar antwoord.","title":"Verzoeken en wachttijd bij piekbelasting"},"worked":{"body":"De chatmeting is uitgesloten omdat de sanity-check is mislukt. De aanwezige tokens/s bewijzen daardoor geen bruikbare modeluitvoer.","title":"Chat bij tien gelijktijdige verzoeken"},"surprised":{"body":"0 van de 11 tests hebben een meting die volgens de huidige Arena-regels meetelt. De sanity-check is mislukt. Deze run wordt daarom uitgesloten van de rangschikking; de ruwe cijfers blijven alleen ter inspectie zichtbaar. Afgebroken: 11-rate-sweep.","title":"Welke metingen meetellen"}},"hardware":"DGX Spark, NVIDIA GB10, 128 GB unified memory","modelUrl":"https://huggingface.co/mistralai/Ministral-3-3B-Instruct-2512","provider":"Mistral AI","precision":"BF16","parameters":"3B","provenance":{"modelId":"mistralai/Ministral-3-3B-Instruct-2512","runPath":"results/ministral-3/ministral-3-3b-instruct/bf16","validity":{"tests":{"01-chat":"complete","02-rag-8k":"complete","08-sharegpt":"complete","09-reasoning":"complete","04-multi-turn":"complete","11-rate-sweep":"aborted","03-long-output":"complete","05-big-context":"complete","10-monday-peak":"complete","07-office-baseline":"complete","06-concurrency-stress":"complete"},"sanity":"failed"},"sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","servedName":"ministral-3-3b-instruct-bf16","generatedAt":"2026-08-09T00:19:16+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":"ministral-3-3b-instruct-bf16-chat-1k","date":"2026-08-09","runs":3,"metrics":{"ttfrSec":{"mean":0.864,"stddev":0.182},"decodeTokensPerSecondTotal":{"mean":356.5,"stddev":9.72},"prefillTokensPerSecondTotal":{"mean":10083.28,"stddev":1230.71},"decodeTokensPerSecondPerUser":{"mean":42.31,"stddev":0.63}},"sourceRun":"Test 01","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","workloadId":"chat-1k","sourceLabel":"mistralai/Ministral-3-3B-Instruct-2512 bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"01-chat","prefixCaching":false},{"id":"ministral-3-3b-instruct-bf16-rag-8k","date":"2026-08-09","runs":3,"metrics":{"ttfrSec":{"mean":3.403,"stddev":1.718},"decodeTokensPerSecondTotal":{"mean":157.9,"stddev":0.94},"prefillTokensPerSecondTotal":{"mean":12475.72,"stddev":22.21},"decodeTokensPerSecondPerUser":{"mean":17.46,"stddev":1}},"sourceRun":"Test 02","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","workloadId":"rag-8k","sourceLabel":"mistralai/Ministral-3-3B-Instruct-2512 bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"02-rag-8k","prefixCaching":false},{"id":"ministral-3-3b-instruct-bf16-long-output-agents","date":"2026-08-09","runs":3,"metrics":{"ttfrSec":{"mean":0.461,"stddev":0.03},"decodeTokensPerSecondTotal":{"mean":238.93,"stddev":69},"prefillTokensPerSecondTotal":{"mean":5008.19,"stddev":195.58},"decodeTokensPerSecondPerUser":{"mean":50.17,"stddev":0.62}},"sourceRun":"Test 03","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","workloadId":"long-output-agents","sourceLabel":"mistralai/Ministral-3-3B-Instruct-2512 bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"03-long-output","prefixCaching":false},{"id":"ministral-3-3b-instruct-bf16-multi-turn-office","date":"2026-08-09","runs":3,"metrics":{"ttfrSec":{"mean":1.21,"stddev":0.305},"decodeTokensPerSecondTotal":{"mean":329.33,"stddev":3.13},"prefillTokensPerSecondTotal":{"mean":12155.42,"stddev":88.83},"decodeTokensPerSecondPerUser":{"mean":34.98,"stddev":0.77}},"sourceRun":"Test 04","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","workloadId":"multi-turn-office","sourceLabel":"mistralai/Ministral-3-3B-Instruct-2512 bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"04-multi-turn","prefixCaching":false},{"id":"ministral-3-3b-instruct-bf16-large-context-25k","date":"2026-08-09","runs":3,"metrics":{"ttfrSec":{"mean":15.135,"stddev":8.037},"decodeTokensPerSecondTotal":{"mean":46.11,"stddev":0.11},"prefillTokensPerSecondTotal":{"mean":8074.45,"stddev":7.34},"decodeTokensPerSecondPerUser":{"mean":6.25,"stddev":1.17}},"sourceRun":"Test 05","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","workloadId":"large-context-25k","sourceLabel":"mistralai/Ministral-3-3B-Instruct-2512 bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"05-big-context","prefixCaching":false},{"id":"ministral-3-3b-instruct-bf16-concurrency-stress-25k","date":"2026-08-09","runs":3,"metrics":{"ttfrSec":{"mean":31.187,"stddev":18.617},"decodeTokensPerSecondTotal":{"mean":44.34,"stddev":0.08},"prefillTokensPerSecondTotal":{"mean":6959.45,"stddev":7.14},"decodeTokensPerSecondPerUser":{"mean":3.17,"stddev":0.71}},"sourceRun":"Test 06","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","workloadId":"concurrency-stress-25k","sourceLabel":"mistralai/Ministral-3-3B-Instruct-2512 bf16 op de DGX Spark","sourceStatus":"complete","sourceTestId":"06-concurrency-stress","prefixCaching":false}],"openLoopResults":[{"id":"ministral-3-3b-instruct-bf16-office-random-4k","date":"2026-08-08","metrics":{"e2eMs":{"p50":17448.68,"p90":33722.99,"p95":39038.81,"p99":41995.52,"mean":18541.8},"tpotMs":{"p50":36.86,"p90":46.99,"p95":50.08,"p99":55.43,"mean":36.85},"ttftMs":{"p50":660.09,"p90":1038.43,"p95":1193.26,"p99":1364.85,"mean":705.35}},"sourceRun":"Test 07","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","scenarioId":"office-random-4k","achievedRps":0.292,"sourceStatus":"complete","sourceTestId":"07-office-baseline","configuredRps":0.3,"prefixCaching":false,"totalRequests":200,"successfulRequests":200,"totalTokenThroughput":1330.93,"peakConcurrentRequests":14},{"id":"ministral-3-3b-instruct-bf16-sharegpt-replay","date":"2026-08-08","metrics":{"e2eMs":{"p50":2543.34,"p90":9188.74,"p95":12283.77,"p99":15620.18,"mean":3959.93},"tpotMs":{"p50":18.42,"p90":19.19,"p95":19.76,"p99":21.42,"mean":18.55},"ttftMs":{"p50":68.81,"p90":386.53,"p95":417.43,"p99":460.28,"mean":116.82}},"sourceRun":"Test 08","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","scenarioId":"sharegpt-replay","achievedRps":0.296,"sourceStatus":"complete","sourceTestId":"08-sharegpt","configuredRps":0.3,"prefixCaching":false,"totalRequests":250,"successfulRequests":250,"totalTokenThroughput":126.77,"peakConcurrentRequests":7},{"id":"ministral-3-3b-instruct-bf16-reasoning","date":"2026-08-08","metrics":{"e2eMs":{"p50":237499.26,"p90":399393.56,"p95":425904.32,"p99":453997.17,"mean":243472.61},"tpotMs":{"p50":60.92,"p90":69.08,"p95":70.26,"p99":70.89,"mean":56.77},"ttftMs":{"p50":422.68,"p90":650.84,"p95":679.49,"p99":761.83,"mean":424.07}},"sourceRun":"Test 09","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","scenarioId":"reasoning","achievedRps":0.08,"sourceStatus":"complete","sourceTestId":"09-reasoning","configuredRps":0.2,"prefixCaching":false,"totalRequests":50,"successfulRequests":50,"totalTokenThroughput":423.34,"peakConcurrentRequests":37},{"id":"ministral-3-3b-instruct-bf16-monday-peak-random-4k","date":"2026-08-08","metrics":{"e2eMs":{"p50":46479.36,"p90":86714.6,"p95":90163.41,"p99":94177.94,"mean":47987.42},"tpotMs":{"p50":97.04,"p90":101.78,"p95":106.25,"p99":113.46,"mean":94.92},"ttftMs":{"p50":803.42,"p90":1229.55,"p95":1340.04,"p99":1792.47,"mean":851.22}},"sourceRun":"Test 10","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","scenarioId":"monday-peak-random-4k","achievedRps":0.501,"sourceStatus":"complete","sourceTestId":"10-monday-peak","configuredRps":1.5,"prefixCaching":false,"totalRequests":300,"successfulRequests":300,"totalTokenThroughput":2252.76,"peakConcurrentRequests":28}],"rateSweep":{"date":"2026-08-09","capacity":{"slo2000":{"outputTps":138.1,"ttftP95Ms":1566.8,"achievedRps":0.281,"configuredRps":0.3,"peakConcurrent":14},"slo5000":{"outputTps":225.4,"ttftP95Ms":2049.8,"achievedRps":0.461,"configuredRps":0.5,"peakConcurrent":37},"slo10000":{"outputTps":225.4,"ttftP95Ms":2049.8,"achievedRps":0.461,"configuredRps":0.5,"peakConcurrent":37}},"sourceRun":"Test 11","sourceUrl":"https://github.com/djangodevreng/dgx-spark-benchmarks/tree/a948cec23b054c62fef595bad3cdb284066922ac/results/ministral-3/ministral-3-3b-instruct/bf16/","abortedAtRps":1,"queueKneeRps":null,"sourceStatus":"aborted","sourceTestId":"11-rate-sweep","totalRequests":425,"successfulRequests":425}}