Native LLM inference server for Apple Silicon. OpenAI + Anthropic API compatible. No Python. Zig backend, Swift frontend macOS app with chat, music, voice, video generation.
C-Rank Grade: A (Welcoming) - 64/100
External PR Merge Rate: 51%
Response Time: -
First Timer Success: 50%
ddalcu/mlx-serve has a recorded first-timer success rate of 50.0%. Repositories ranked A typically provide actionable feedback during code reviews and actively nurture new community contributors.
Maintainers in ddalcu/mlx-serve respond to incoming external pull requests in approximately 0.0 hours on average. Keeping PRs focused on single tasks and ensuring tests pass helps maintainers review faster.
The C-Rank™ system evaluates GitHub projects on a 0–100 scale using real data: PR merge rates, review turnaround time, active maintainer presence, and first-time contributor success. A score of 64.1 places ddalcu/mlx-serve in the A tier.
The external contributor pull request merge rate for ddalcu/mlx-serve is 50.9%, based on public PR activity from non-core contributors.
ddalcu/mlx-serve currently has 2 active issue(s) tagged with beginner-friendly labels like "good first issue", "beginner", or "up-for-grabs".
Highly welcoming maintainers in ddalcu/mlx-serve. Prompt code reviews with positive guidance for new contributors.
Average Response Latency
Tracks hours until a maintainer leaves a review, comment, or PR response.
External Acceptance Rate
Percentage of community pull requests successfully merged into main.
First PR Conversion
Rate at which developers submitting their first repository PR succeed.
Show contributors that your repository actively reviews and merges external pull requests.
[](https://getmerged.abhishekco.de/ddalcu/mlx-serve?utm_source=github&utm_medium=badge)
Historical health & responsiveness evolution
External PR acceptance
First PR success rate
50% receive response in
90% receive response in
Maintainer review activity heatmap & prime window prediction
Calculating local time...
Calculating local time...
Detected: UTC
Maintainers are most responsive around 01:00 UTC (01:00 UTC in your timezone). Recommended submission window is Wednesday - Friday from 00:00 to 04:00 UTC (00:00 - 04:00 UTC local).
Maintainer Strategy: Maintainer Sweet Spot: PRs under 150 lines merge in ~16 hours (64% approval). Large PRs (>500 lines) take ~115 hours (23% approval). Keep your diff compact for rapid review.
Setup: mlx-serve 26.8.11 (Homebrew tap), M3 Ultra 96 GB, MiniMax-H3-FL2VA-MLX-Serve-8bit + MiniMax-H3-REF2VA-MLX-Serve-8bit, driving /v1/video/generations over HTTP. What I'm trying to do: clips longer than one window (~10-15 s+) that keep the identities from ref_images — the "family clip" use case: a few reference photos, one scene prompt, 20-30 s out. The caller is a video-generation tool wired into Hermes Agent (which your launch command already knows about), so server-side chaining matters more than it would for a hand-driven client. What happens: chain_windows > 1 together with the REF2VA pack returns: 400: chained windows ride FL2VA keyframe conditioning — the REF2VA pack cannot serve them; load an FL2VA checkpoint That gate makes sense given the two checkpoints. My question before anything else: am I missing a supported way to do this today? Things I considered from the docs/source: Manual continuation via ref_videos: generate window N, subsample its frames, pass them (plu
Summary When mlx-serve dies unexpectedly, MLX Core shows a modal alert and stays stopped. There is no recovery: the server remains down until a human notices the modal and restarts it manually. On a Mac Studio serving the LAN (coding agents, other Macs), a crash takes down every client until someone is at the keyboard. What happens today app/Sources/MLXServe/Services/ServerManager.swift: handleTermination(exitCode:) — when the process exits while .running, it sets .error("Exited unexpectedly") and calls presentCrashAlert(...). presentCrashAlert — a modal NSAlert.runModal() with "Copy Log" / "OK". Dismissing it leaves the server stopped. Intentional stops are already distinguished: stop() nils proc.terminationHandler before terminate(), so only an unexpected exit reaches the alert path. There is no restart anywhere in the app; the only "restart" concept is the manual "Restart Now" banner for settings changes. The building blocks for recovery already exist: lastLaunchedOptions —
Would it be possible to add an optional idle timeout for loaded models? For example, after 10 minutes without requests, the model would be unloaded and its MLX/Metal memory released. The next request would load it again. This would be useful on Macs with limited unified memory, especially when running large models. For example https://github.com/ggml-org/Llama-macOS supports it via Thanks!
Rate what actually matters after opening a pull request here.
Rate your first-hand PR experience (review speed, maintainer responsiveness, and onboarding ease) to help other contributors.
When evaluating whether to contribute to ddalcu/mlx-serve, response velocity and maintainer engagement are crucial. GetMerged continuously tracks pull request trajectories, first-comment latency, and code review rounds to help developers avoid submitting pull requests to backlogged repositories.
Currently, maintainers of ddalcu/mlx-serve acknowledge new external contributions in approximately -. Out of all submitted pull requests from non-core authors in the last 180-day window, 50.9% were successfully merged into the primary branch.
ddalcu/mlx-serve has a recorded first-timer success rate of 50.0%. Repositories ranked Welcoming typically provide actionable feedback during code reviews and actively nurture new community contributors.
The initial maintainer response time averages ~-. Keeping PRs scoped to single concerns and ensuring CI checks succeed will optimize review turnaround.
The C-Rank™ index scores repositories on a 0 to 100 scale using an objective formula: external PR merge rates, initial response speed, active maintainer count, and first-time contributor retention. A score of 64.1 places ddalcu/mlx-serve in the Welcoming tier.
The external pull request merge rate is 50.9%. GetMerged isolates non-core community contributions so external developers get an accurate benchmark of PR acceptance probability.
Yes, ddalcu/mlx-serve currently has 2 active issue(s) tagged with beginner-friendly labels. You can inspect these directly from the repository issues tab.
All metrics displayed for ddalcu/mlx-serve are automatically retrieved via the public GitHub API and recalculated daily. Insider pull requests submitted by repository owners or organization members are excluded from merge rate calculations to preserve objective external contributor statistics.