Score breakdown
Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.
Why it matters
Useful for Apple Silicon developers who want a single-binary H3 inference path without a Python stack, researchers studying on-device video/audio inference performance on M3 Max / M5 Max, and open-source generative-video tooling maintainers tracking the smallest end-to-end stack.
Who should use it
Who should skip it
Skip antirez/h3.c unless the captured evidence suggests it solves a problem you are actively working on.
About this signal
antirez/h3.c is tracked by RepoRadar as a framework in the On-Device Inference section. First seen 2026-08-11; the source record was last checked on 2026-08-11. The current verdict is 'try now' with a Gold tier and moderate setup difficulty. antirez/h3.c leads on novelty (10.0) and workflow potential (10.0); its lowest signal is setup ease (6.4), so factor that in before investing setup time. This page summarizes the public evidence on the linked source page and states where additional review is still needed.
How this item is evaluated
The antirez/h3.c record combines a 9.0/10 composite score with separate popularity (7.4), risk (conditional), and setup (moderate) signals. See the scoring methodology for the current weights and evidence definitions.
Questions worth asking before you adopt this
Putting this into practice? Read How to evaluate an AI tool before you adopt it for the checklist behind this score.
Risk explanation
The h3.c engine is a C + Metal implementation that maps BF16 prompt conditioning, prepared DiT, and video decoder in memory; --show adds roughly 10 GiB of temporary model residency, so size unified memory before running with --show on M3 Max / M5 Max.