Claude Fable 5.1 & GPT-6 Astra packages are live

Performance

15 packages, each available for 11 model families.

budgets

Performance budgets as a CI contract — p50/p95 latency and RSS caps per scenario, a reproducible benchmark script, a committed baseline, and a…

11 models

bundle-size

Keeping JavaScript small — measuring before cutting, dependency discipline, tree shaking that actually works, and budgets enforced in CI.

11 models

caching

Caching layers, invalidation strategies, stampede protection and the correctness rules that keep a cache from serving one user another user's data.

11 models

cpu

CPU-bound work — profiling with flame graphs, algorithmic complexity, keeping event loops free, parallelism, and knowing when CPU is not the problem.

11 models

database

Database performance at the system level — connection pooling, saturation, lock contention, bloat, and the signals that tell you which one you have.

11 models

fonts

Web font delivery — self-hosting, subsetting, font-display, metric-compatible fallbacks, and eliminating the layout shift a font swap causes.

11 models

go-performance

Making Go fast with evidence — pprof first, allocations and escape analysis, preallocation, strings.Builder, sync.Pool where it pays, GC tuning…

11 models

images

Image delivery — modern formats, responsive sizes, reserved space, lazy loading, and the LCP image that must never be deferred.

11 models

lazy-loading

Deferring work until it is needed — components, images, data and third-party scripts — without creating waterfalls or hurting the metrics you meant…

11 models

memory

Finding and fixing memory problems — leaks versus growth, heap snapshots, streaming instead of buffering, and sizing a runtime inside a container.

11 models

network

Network performance — latency versus bandwidth, connection setup cost, compression, request waterfalls, and delivering bytes from close to the user.

11 models

optimization

A method for making software faster — measure, profile, find the bottleneck, change the complexity not the constant, and prove the improvement.

11 models

prefetching

Loading things before they are needed — resource hints, prefetch on intent, speculation rules, and not wasting a user's bandwidth or battery.

11 models

queries

Query performance from the application's side — N+1 elimination, projection, batching, pagination cost, and asserting query counts in tests.

11 models

rendering

Rendering performance — the frame budget, layout thrash, compositor-only animation, virtualised lists, and keeping interaction responsive.

11 models

Other categories