{"data":{"sourceKind":"published_summary","id":"4aeec25f-3a0f-53c1-bd54-df12cd828bd0","producerId":"inferencex","sourceUrl":"https://inferencex.semianalysis.com/api/v1/benchmarks?model=DeepSeek-R1-0528&runId=25980019087&exactRun=true","sourceRepository":"https://github.com/SemiAnalysisAI/InferenceX","observedAt":"2026-08-15T01:05:11.178Z","model":"dsr1","scenario":"single_turn, 2048 total tokens, concurrency 4","precision":"fp8","interactivityTokensPerSecondPerUser":103.09633817958112,"primaryMetric":"throughput_tokens_per_second_per_gpu","canonicalUnit":"tokens/second/GPU","baseline":{"ecosystem":"rocm","hardware":"mi355x","value":92.52648660357187},"candidate":{"ecosystem":"cuda","hardware":"h200","value":98.82861137108823},"relativeDelta":0.06811157538616705,"comparability":"exact_configuration","rawSamplesAvailable":false,"materialChangeEligible":false,"caveat":"InferenceX publishes this exact configuration summary and workflow provenance without raw samples. Upstream treats the comparison as contextual and never as a material change.","evidence":{"evidenceId":"4aeec25f-3a0f-53c1-bd54-df12cd828bd0","sourceType":"benchmark","title":"mi355x and h200 InferenceX comparison","upstreamUrl":"https://inferencex.semianalysis.com/api/v1/benchmarks?model=DeepSeek-R1-0528&runId=25980019087&exactRun=true","sourceTime":"2026-05-27T00:00:00.000Z","sanitizedSha256":"c07bf618517a56c719338759f23ba191ecbd97e182d3115d9fe52b782a1cb4e6","freshness":"fresh"},"sourceObservations":[{"providerResultId":"415860","sourceUrl":"https://inferencex.semianalysis.com/api/v1/benchmarks?model=DeepSeek-R1-0528&runId=26486957776&exactRun=true","sourceSha256":"da7cfb59d6d139a58ee74252e106f1bb03a91dfcd400c4bd271ba4e930fdf547","workflowRunId":"26486957776","workflowAttempt":1,"workflowHeadSha":"7a8a5ab133f20911b7f43d4e70787890e5dfa293","sourceCommitSha":"7a8a5ab133f20911b7f43d4e70787890e5dfa293","workflowUrl":"https://github.com/SemiAnalysisAI/InferenceX/actions/runs/26486957776/attempts/1","imageReference":"lmsysorg/sglang:v0.5.12-rocm700-mi35x"},{"providerResultId":"409936","sourceUrl":"https://inferencex.semianalysis.com/api/v1/benchmarks?model=DeepSeek-R1-0528&runId=25980019087&exactRun=true","sourceSha256":"c07bf618517a56c719338759f23ba191ecbd97e182d3115d9fe52b782a1cb4e6","workflowRunId":"25980019087","workflowAttempt":2,"workflowHeadSha":"2caa27420cd842f0caf8f1f0032458c94bb4f445","sourceCommitSha":"2caa27420cd842f0caf8f1f0032458c94bb4f445","workflowUrl":"https://github.com/SemiAnalysisAI/InferenceX/actions/runs/25980019087/attempts/2","imageReference":"lmsysorg/sglang:v0.5.12-cu130"}],"capabilityMappingBasis":"missing","releaseMappingBasis":"missing"},"page":null,"meta":{"asOf":"2026-08-15T01:05:11.178Z","snapshotId":null,"contextSnapshotId":"5453ed9b-88df-5536-b274-e53c3497dadc","collectionRevision":"3462ea9a368f360ed09abf4a0a38d957a3cbc2fef3950f31857b32785c010395","membershipBasis":"live_collection","sourceCutoff":null,"freshness":"unavailable","freshnessPolicyVersion":"freshness-v1","taxonomyVersion":"513f67b1-9ea7-46cc-b713-d6ea30ffb93f","metricVersions":{"ci":"ci-health:v2","contribution":"contribution-health:v2"},"cohortVersion":"aac24f2b-1109-5deb-a608-1c3d55fb0ae6","queryFingerprint":"ee0ba8c2880b9628a7e5f922e70a14542b29d5dd9a3e74e8f9b24372f91d61ab","sourceMode":"live","apiVersion":"v1"}}