A private, offline alternative to cloud AI like Claude and ChatGPT. Outlier runs open models — up to 397B parameters — locally on Apple Silicon, so there are no caps, no data centers, and nothing leaves your device.
v1.11.756 · notarized Mac DMG · Apple Silicon (M1–M4)
Here's the middleware rewritten with async/await and clean error propagation:
export async function auth(req, res, next) { try { const token = req.headers.authorization?.split(' ')[1]; if (!token) return res.status(401).json({ error: 'No token' }); req.user = await verify(token); next(); } catch (err) { next(err); // forward to the error handler } }
Generated locally on your Mac — nothing left your device
Local model inference runs entirely on your Apple Silicon chip. Your prompts, files, code, and chat history stay on your disk, under your control. Only model downloads and optional web search ever touch the network — and only when you choose. No telemetry on inference. The open-weight models are auditable on HuggingFace.
Today's build is the foundation: local chat, session history, a model picker with on-device downloads, project context, memory, web research, agent tools with approval, and vision — in a signed, notarized Mac app.
Streaming token output, persistent local history, rename / delete / pin, Markdown export — and a running cost display that stays at $0.00.
File read / write and shell execution behind permission modes, a plan-review card, repair loop, audit log, path scoping, project map, and tests.
A research mode over DuckDuckGo with a Wikipedia fallback, source filters, summary cards, trust badges, and inline citations with source excerpts.
Persistent memory in SQLite with short / medium / long-term tiers, provenance tracking, review cards, conflict detection, decay, and MEMORY.md export.
Upload an image or a screenshot and query it directly through Outlier Vision — built for OCR, diagrams, and multimodal Q&A.
A macOS arm64 DMG, accepted by Apple notarization and Gatekeeper, distributed via GitHub Releases with a built-in auto-updater.
Free is useful immediately with Nano + Lite. Pro ($20/mo or $149/yr) adds the other five tiers — Quick, Core, Code, Vision, and the 397B Plus tier. The lineup below reflects the current shipping models.
| Tier | Plan | Best for | Disk / RAM | Speed / note |
|---|---|---|---|---|
Outlier Nano Qwen3.5-4B · MLX 4-bit |
Free | Fast iteration, lightweight chat, small Macs | 2.37 GB · 6 GB min RAM | ~30–70 tok/s (Mac-dependent) |
Outlier Lite Qwen3.5-9B · MLX 4-bit |
Free | Daily local AI, writing, search, Q&A | 5.04 GB · 12 GB min RAM | Mac-dependent |
Outlier Quick Gemma-4-26B MoE |
Pro | Thinking-mode reasoning, not a code substitute | 15.61 GB · 16 GB min RAM | Mac-dependent |
Outlier Core Qwen3.6-27B · text-only |
Pro | Best default quality, reasoning, coding | 15.13 GB · 24 GB min RAM | High-end quality, fully offline |
Outlier Code Core weights + code config |
Pro | Coding workflow, lower-temp code-tuned setup | 15.13 GB · 24 GB min RAM | Same verified base as Core |
Outlier Plus Qwen3.5-397B-A17B · V9 paged |
Pro | Largest local tier · 397B MoE on high-end Macs | 209 GB disk · 64 GB min, 128 GB recommended | Trophy tier — capability over speed |
Outlier Vision Qwen3.6-35B-A3B · vision retained |
Pro | Images, screenshots, OCR, multimodal reasoning | 19.0 GB · 24 GB min RAM | V9 K=256: 16.31 tok/s @ 34.04 GB |
Quick / Core / Code / Vision / Plus 397B are all included with Pro ($20/mo or $149/yr) or lifetime Pro. Code uses the same weights as Core with code-specialized configuration. Quick is useful for reasoning, not positioned as a coding tier.
Every tier is sized to real Apple Silicon memory. Start free on a MacBook Air; scale all the way to the 397B Plus tier on a Mac Studio.
Apple Silicon only (M1 / M2 / M3 / M4). Intel Macs are not supported.
The cloud tools proved the workflow: coding agents, long-context research, always-on help. The problem is the meter. Outlier is building the local version — Mac-native, private by default, and not capped by an Outlier token allowance.
| Cloud AI tools | Cloud coding agents | Outlier Free | Outlier Pro | |
|---|---|---|---|---|
| Monthly cost | $20+ | Often much higher | $0 | $20/mo · $149/yr |
| Usage model | Server-side limits | Usage windows / caps | No token meter | No token meter |
| Where inference runs | Provider cloud | Provider cloud | Your Mac | Your Mac |
| Privacy default | Remote request | Remote repo / context | Local by default | Local by default |
| Offline use | No | No | Yes, once downloaded | Yes, once downloaded |
| Current maturity | Very mature | Very mature | Useful beta | Ambitious beta |
Honest framing: Outlier is not claiming parity with the best cloud coding agents today. The beta is the foundation; Pro and Founders revenue funds the climb toward that experience, locally. See the raw data: 54-prompt Outlier vs Claude · Mac local-AI benchmarks · streaming-engine tok/s · RAM → model size.
No investors and no API margins to protect — that's why Free is genuinely free, Pro is $20/mo, and lifetime starts at $99. Pro includes everything: all seven tiers, Plus 397B, and every feature.
Local, private AI for everyday use — no subscription required.
Every model and every feature Outlier ships, including the 397B Plus tier.
Pay once, own Pro forever — the lowest price local Pro will ever be.
14-day money-back guarantee — just email matt@outlier.host. Prefer to back the build without a subscription? Chip in any amount or email Matt to sponsor a benchmark run.
Cloud inference needs data centers, networking, cooling, and ever-growing GPU clusters. Local inference uses the Apple Silicon chip you already own — no round-trip, no per-token meter.
of AI compute is inference — the token calls made every time someone uses a cloud model, now drawing more grid electricity than training did (MIT Tech Review, 2025). Outlier runs them on the Mac you already own.
Local models create no Outlier cloud-inference bill and no cloud token meter.
Apple Silicon unified memory is efficient for local inference versus shipping every prompt to a server.
This is not a claim that every local query is automatically cleaner in every situation. Hardware, model size, electricity source, and usage pattern all matter. The point is directional: if a large share of everyday AI inference moves from data centers to efficient devices people already own, the load on cloud infrastructure can drop.
That's why compression, routing, quantization, and paging aren't just engineering details — they're part of the product philosophy. A useful local model isn't only cheaper for you; it can reduce unnecessary cloud dependence for everyday tasks.
The best environmental feature isn't a green badge. It's a model that's good enough, small enough, and fast enough that people actually choose to run it locally.
Outlier/restart if it wedges.Cloud tools proved what the workflow should feel like. Outlier is the version that runs locally, belongs to you, and never hits a token cap mid-task.
Requires a Mac with Apple Silicon (M1, M2, M3, or M4) — Intel Macs are not supported. macOS 12+.