01 / 66 pages
Open Weights, Serving Decisions
Compare model families through weight residency, attention state, numerical formats and deployment requirements. Includes a dated review of recent open-weight releases.
Premium PDFKeep the reasoning.
Take it with you.
Five detailed references for the decisions between a model checkpoint and a reliable stream of tokens. Worked calculations, technical drawings and source notes, ready to read away from a browser.
5 reference booksPDF · searchable text · linked contentsOctober 2026 editions
02 / 64 pages
Accelerators for Token Serving
A hardware reference organized around serving decisions: memory per rank, interconnect limits, software support and useful throughput under a latency target.
Free with an account
03 / 106 pages
The Kubernetes Inference Runtime
Follow cache-aware routing, prefill and decode pools, multi-node workers and rollouts through a Kubernetes serving graph, with worked capacity and failure cases.
Premium PDF
04 / 117 pages
Operating the Token Path
An operating manual for engine configuration, latency clocks, memory limits, profiling, observability and incident recovery. Includes a CPU analysis companion.
Premium PDF
05 / 77 pages
The Inference Engineer’s Lexicon
A linked reference from GPU execution and memory traffic to distributed decoding and service metrics. Worked cases expose the distinctions that matter in inference reviews.
Premium PDFA SHORTER START
The original inference field guide
Eight pages of serving decisions and worked calculations. Read the online edition with Premium, or download it with eligible PDF access.
Open the compact field guide →READ WITH THE WORKLOAD IN VIEW
Check the edition, then test the assumption.
Each book states its source review dates. Specifications and engine behavior can change. Worked examples explain the arithmetic; the books do not claim new GPU benchmarks.
Work through the inference lab →