rapid-mlx 0.11.5

Rapid-MLX — AI inference for Apple Silicon. Drop-in OpenAI API, 2-4x faster than Ollama.

The fastest local AI engine for Apple Silicon.
Drop-in OpenAI / Anthropic API · 2–4× faster than Ollama · Runs on any M-series Mac.

_{rapidmlx.com · Docs · Model mirror · Desktop app}

---

Quick Start (60 seconds)

1. Install — pick one path (run only one of these):

One-liner — detects your RAM, picks a starter model (recommended):

curl -fsSL https://rapidmlx.com/install.sh | bash

or Homebrew — prebuilt bottle straight from homebrew-core:

brew install rapid-mlx

Both land the same rapid-mlx CLI. The curl installer additionally installs Python 3.10+ if missing, creates an isolated venv at ~/.rapid-mlx/, symlinks the rapid-mlx CLI into ~/.local/bin/, and prints a serve command sized to your Mac (8–23 GB → qwen3.5-4b-4bit; 24–47 GB → gpt-oss-20b-mxfp4-q8; 48–95 GB → qwen3.6-35b-8bit; 96 GB+ → gpt-oss-120b-mxfp4-q8).

Install security. install.sh is served over HTTPS (HSTS-preload) from rapidmlx.com and is a byte-identical mirror of install.sh at the release commit — read it before running if you like. If you want a cryptographically verified installer rather than trusting the website pipe, don't curl | bash the URL above: instead download the release's install.sh asset, verify it against the cosign-signed SHA256SUMS.txt shipped alongside it, and run that verified copy — full recipe in SECURITY.md. PyPI artifacts additionally carry Sigstore attestations (PEP 740). Two more low-trust paths:

• Pin to a commit hash — curl -fsSL https://raw.githubusercontent.com/raullenchai/Rapid-MLX/<commit>/install.sh -o install.sh && shasum -a 256 install.sh && bash install.sh
• Skip the shell script entirely — use Homebrew, uv, or pip below.

See Alternative install methods for the non-curl paths.

2. Chat with a model right now:

rapid-mlx chat

Defaults to qwen3.5-4b-4bit. First run downloads the weights (~2.5 GB) with a progress bar and drops you into a REPL. Type /help for slash commands, /exit to quit.

3. Or serve it for use from other apps:

rapid-mlx serve qwen3.5-4b-4bit

Starts an OpenAI-compatible HTTP server bound to http://localhost:8000. Point any OpenAI SDK / client (Cursor, Aider, LangChain, OpenCode, PydanticAI, your own scripts) at http://localhost:8000/v1; Claude Code / Anthropic SDK uses http://localhost:8000 (the Anthropic messages route lives at /v1/messages under the same host).

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"default","messages":[{"role":"user","content":"Say hello"}]}'

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8000/v1", api_key="not-needed")
print(client.chat.completions.create(
    model="default",
    messages=[{"role": "user", "content": "Say hello"}],
).choices[0].message.content)

4. Or wire up your coding agent — one command:

rapid-mlx launch claude-code

With a server running (step 3), this patches Claude Code's local config (~/.config/claude/settings.json) to route at http://localhost:8000 — no manual env vars, no editing JSON by hand. You get a fully local Claude Code: $0 per token, nothing leaves your Mac. Swap in cursor, cline, or continue-dev for the other IDE clients, or run rapid-mlx launch list to see what's detected on this machine.

Vision / audio / video / diffusion models? Base install is text-only (~460 MB). Vision, audio (TTS, STT, voice cloning), video generation, embeddings, and DFlash speculative decoding ship as opt-in extras. → Optional extras

Not into the terminal? Rapid-MLX Desktop bundles the same engine inside a one-click Mac app.

---

Video generation

Run text-to-video or image-to-video locally through the OpenAI-compatible Videos API. Three backends ship — Wan 2.1 / 2.2, CogVideoX-Fun and LTX-2.3 — across 8 registered checkpoints. wan2.2-ti2v-5b-q8 is the recommended starting point: smallest of the Wan set, and TI2V means one checkpoint does both text-to-video and image-to-video.

Requires Python 3.11+ (the video runtime does not support 3.10; core text and audio still do) and ffmpeg for the final MP4 mux.

pip install 'rapid-mlx[video]'
brew install ffmpeg
rapid-mlx serve wan2.2-ti2v-5b-q8

Create and download a clip:

curl http://localhost:8000/v1/videos \
  -F model=wan2.2-ti2v-5b-q8 \
  -F 'prompt=A fox running through fresh snow, cinematic tracking shot' \
  -F seconds=1 \
  -F size=832x512

# Poll until GET /v1/videos/VIDEO_ID reports "status": "completed", then:
curl http://localhost:8000/v1/videos/VIDEO_ID/content -o output.mp4

The create call returns a job immediately. Poll GET /v1/videos/VIDEO_ID until status is completed. Add -F input_reference=@start.png for image-to-video.

Generation is serialized — one clip at a time — because two diffusion pipelines resident at once will exhaust unified memory. Expect minutes of compute per second of footage, not real time.

→ Every checkpoint, RAM requirement and tuning knob

---

Audio: speech, transcription, voice cloning

41 audio aliases behind the OpenAI-compatible /v1/audio/* endpoints — any OpenAI SDK works unchanged.

pip install 'rapid-mlx[audio]'

# Text to speech
rapid-mlx serve kokoro
curl http://localhost:8000/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"model":"kokoro","input":"hello from rapid-mlx"}' --output hello.wav

# Transcription (Whisper / Parakeet / SenseVoice)
rapid-mlx serve whisper-large-v3-turbo
curl http://localhost:8000/v1/audio/transcriptions \
  -F file=@hello.wav -F model=whisper-large-v3-turbo

Beyond the basics, three things you may not expect to run locally:

• Zero-shot voice cloning from a reference clip. indextts is the only one that takes the clip alone; qwen3-tts-clone, f5-tts-zh and chatterbox all require ref_text (the clip's exact transcript) paired with ref_audio, and the request is rejected before generation if it is missing.
• Voice design — qwen3-tts-voicedesign has no named speakers at all. Describe the voice you want in natural language via instructions (timbre, gender, age, accent, emotion, prosody) and it synthesises it.
• Forced alignment — qwen3-aligner takes audio plus the transcript you already have and returns per-character timings. It never guesses at the words, so it cannot mis-hear them; that is what karaoke captions and beat-synced editing need.

Also: word-level timestamps on transcription, and local text-to-music at /v1/audio/music.

→ All 41 aliases across 12 families

---

Why Rapid-MLX

Apple-Silicon-native	Pure MLX kernels — no llama.cpp fallback, no Metal shim. Continuous batching, prompt cache (radix + DeltaNet RNN snapshots), and a quantized live KV cache (int4/int8 on the continuous-batching cache + TurboQuant K8V4 codec) run at native MLX bandwidth on M1 → M4.
Drop-in OpenAI / Anthropic API	/v1/chat/completions, /v1/responses (Codex CLI), /v1/messages (Anthropic SDK / Claude Code), /v1/embeddings, /v1/audio/*, /v1/videos — same wire as ChatGPT / Claude, no client adapter.
First-class ecosystem coverage	11 agent CLIs and 3 Python frameworks are wire-verified against real weights every release (4 are Tier-1, re-verified on current binaries) — Codex CLI, Claude Code, OpenCode, Qwen Code, OpenHands, Hermes Agent, Aider, Kilo Code, GitHub Copilot, Factory Droid, Moonshot Kimi Code + LangChain, PydanticAI, smolagents.

→ Full feature breakdown

---

Use Cases

Chat in the terminal	rapid-mlx chat qwen3.5-9b-4bit	Streaming REPL, /help for slash commands, --think / --no-think to control CoT.
OpenAI server for your apps	rapid-mlx serve qwen3.5-9b-4bit	Point Cursor, Aider, LibreChat, Open WebUI, LangChain at http://localhost:8000/v1.
Agent backends	rapid-mlx serve qwen3.6-35b-8bit & rapid-mlx agents codex --setup && codex	8 agents auto-configure via agents <name> --setup once the server is up (11 wire-verified total, 4 Tier-1) — see Agent support.
Benchmark your Mac	rapid-mlx bench qwen3.5-9b-4bit --submit	Standardized B=1 bench, opens a PR to publish your row on rapidmlx.com.

→ One-shot IDE setup with rapid-mlx launch <cursor|claude-code|cline|continue-dev>

---

Agent Support

All 11 agents below are wire-verified against real weights every release via their own integration-test cell. Of these, four are Tier-1 — Claude Code, Codex CLI, Hermes, and Aider — re-verified end-to-end against the current client binary every release, with one guardian per API wire (Anthropic /v1/messages, OpenAI /v1/responses, and /v1/chat/completions covered for both tool-calling depth and reach). The other seven are Tier-2: wire-verified in the matrix and configured on-demand. The first eight agents each ship a rapid-mlx agents <name> --setup config template (except Claude Code, which is one env-var); GitHub Copilot, Factory Droid, and Moonshot Kimi Code plug in through their own documented BYOK config (auth-gated, so the matrix cell is a wire smoke).

Tier-1 (4): Claude Code · Codex CLI · Hermes · Aider — last re-verified end-to-end 2026-07-28 on current binaries (claude 2.1.211, codex 0.145.0, hermes 0.9.0, aider 0.86.2) against rapid-mlx 0.11.1. Tier-2 (7): OpenCode · Qwen Code · OpenHands · Kilo Code · GitHub Copilot · Factory Droid · Moonshot Kimi Code.

Agents (11)	Frameworks (3)
Codex CLI · Claude Code · OpenCode · Qwen Code · OpenHands · Hermes Agent · Aider · Kilo Code · GitHub Copilot · Factory Droid · Moonshot Kimi Code	LangChain (+ LangGraph) · PydanticAI · smolagents

Also compatible with any OpenAI-compatible client via http://localhost:8000/v1 — Cursor, LibreChat, Open WebUI, and more plug in with a single URL change.

→ Full 11-agent + 3-framework matrix (test cells + xfail reasons) → Codex CLI · Claude Code · OpenCode · Qwen Code · OpenHands · Hermes · Aider · Kilo Code · Copilot · Droid · Kimi Code

---

Choose Your Model

The installer's RAM detector picks a sensible default. If you want to shop the full catalog: rapid-mlx models lists every alias, rapid-mlx info <alias> shows the per-alias profile (parser, MoE / hybrid flags, KV codec eligibility, speculative-decoding gates).

RAM	Recommended	One-shot
8–23 GB MacBook Air/Pro	qwen3.5-4b-4bit	rapid-mlx serve qwen3.5-4b-4bit
24–47 GB MacBook Pro / Mac Mini	gpt-oss-20b-mxfp4-q8	rapid-mlx serve gpt-oss-20b-mxfp4-q8
48–95 GB Mac Studio	qwen3.6-35b-8bit	rapid-mlx serve qwen3.6-35b-8bit
96 GB+ Mac Studio / Pro	gpt-oss-120b-mxfp4-q8	rapid-mlx serve gpt-oss-120b-mxfp4-q8

→ Full RAM tier map + serve flags per tier → Every alias, quant, and family (165 text + 41 audio + 8 video aliases, 214 total) · interactive at models.rapidmlx.com

---

Alternative install methods

The two paths above cover most users — reach for these only if you already manage Python yourself.

Homebrew — Mac-native, one command, prebuilt bottle from homebrew/core

brew install rapid-mlx

Ships in homebrew-core since 0.10.12 — no tap, no trust prompt. Upgrade with brew upgrade rapid-mlx. If you previously installed from the legacy raullenchai/rapid-mlx tap, switch once: brew uninstall rapid-mlx && brew untap raullenchai/rapid-mlx && brew install rapid-mlx.

uv — isolated tool install, auto-manages Python

uv tool install rapid-mlx@latest

Don't have uv yet? curl -LsSf https://astral.sh/uv/install.sh | sh. Upgrade with uv tool upgrade rapid-mlx.

pip — requires Python 3.10+ (macOS ships 3.9)

python3.12 -m pip install rapid-mlx

If pip install rapid-mlx says "no matching distribution", your Python is too old. brew install python@3.12 first. Upgrade with pip install -U rapid-mlx.

For image-input / VLM models (Qwen-VL, true multimodal), install the vision extra: pip install 'rapid-mlx[vision]' — see Optional extras.

---

Command Reference

rapid-mlx --help                    # top-level command list
rapid-mlx <subcommand> --help       # per-subcommand flags

Covers chat, serve, share, agents (setup / test), bench, models, pull, rm, ps, info, doctor, upgrade, telemetry, launch, and jlens.

→ Full CLI reference with every flag

---

Troubleshooting

Run the built-in self-check first:

rapid-mlx doctor

Top three things that go wrong:

• Much slower than expected. Qwen3.5 / 3.6 default to thinking-on — add --no-think to skip chain-of-thought. → Slow tok/s
• Out of memory. Model too big for your RAM — pick a smaller quant from Choose Your Model or the full tier map. → OOM guide
• Tool calls arriving as plain text. Auto-recovery handles most cases; if not, set --tool-call-parser explicitly for your model. → Tool-call recovery

→ All troubleshooting entries (OOM, empty responses, slow TTFT, port taken, shell completion, HF cache, and more)

---

See it in action

Community & Contributing

• Report a bug or request a model: Issues
• Report a security issue: Private advisory — see SECURITY.md
• Ask a question or share a build: Discussions
• Contribute code, aliases, or docs: CONTRIBUTING.md
• Add your hardware to the public benchmark: rapid-mlx bench <alias> --submit opens the PR for you

Rapid-MLX ships opt-in anonymous telemetry (off by default; explicit rapid-mlx telemetry enable required). No prompts, completions, paths, IPs, or API keys are ever collected. → What we do and don't collect

🚀 Contributors

Every avatar here shipped something in rapid-mlx — model support, tool-call parsers, fixes, docs, and benchmark submissions. Thank you.

Star History

---

License

Apache 2.0 — see LICENSE.