RepoRadar
Leave a Tip
Item detail

SemiAnalysisAI/InferenceX

InferenceX is an Apache-2.0 open-source continuous inference benchmark from SemiAnalysis that measures throughput and latency across the major frontier model/hardware combinations that actually ship to production: Kimi K2.6, DeepSeek v4, and GLM-5 on GB200 NVL72, MI355X, B200, and GB300 NVL72, with TPUv6e/v7 and Trainium2/3 on the roadmap. It is the first benchmark to treat the hardware-and-model

Open source →Back to radar
Score7.5
Popularity65.0
Risknone
TierGold
Score breakdown
Usefulness7.0
Novelty8.0
Momentum7.0
Maturity7.3
Open-source/build8.4
Evidence7.2
Workflow potential8.6
Setup ease6.4

Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.

Why it matters

Useful for inference platform teams and procurement engineers who need to defend a hardware decision in a deck: run InferenceX against the model and hardware pair you are about to commit to, share the published numbers, and pair them with at least one real workload before signing the PO.

Who should use it

inference platform teams sizing hardwareprocurement engineers defending a GPU SKU decisionAI infrastructure researchersfounders pitching investors on hardware cost claims

Who should skip it

Skip if the source link, docs, or setup requirements do not match your workflow.

Risk explanation

benchmark numbers are an input, not a substitute for a production pilot; pair InferenceX with at least one real workload before making a hardware commitment.

Evidence links

Closest alternatives / related signals

benchmarkinferencehardwarefrontier-modelssemianalysisinferencexkimideepseek