Use the Code mode in Outlier to scan a PR diff and surface issues without round-tripping to a hosted API. The whole sequence below stays on the Mac.
Apple Silicon Mac, macOS 12 or later, the unified-memory minimum that the chosen tier requires (6 GB for Nano, 12 GB for Lite, 16 GB for Quick, 24 GB for Core and Vision 3.8, 64 GB for Plus). Internet is required only for the one-time model download.
git show HEAD or fetch the PR locally.For PR review, Core sizes its context window to the memory your Mac has free, rather than to a fixed 32K, which matters for long diffs. Both share the same weights, so answer quality is identical.
Large diffs are the limiting factor, not the review itself. Split anything past a few hundred lines by file and review each in its own turn; a diff that overruns the context window gets silently truncated, and a review of half a change reads exactly like a review of all of it. Asking for one axis per pass — correctness, then style, then performance — also keeps each answer short enough to check.
Cloud PR review tools transmit the diff. Outlier’s Code mode keeps it on the Mac, which matters when the diff has untested mitigations or unreleased features.
For PR review, Core sizes its context window to the memory your Mac has free, rather than to a fixed 32K, which matters for long diffs. Both share the same weights, so answer quality is identical.
Large diffs are the limiting factor, not the review itself. Split anything past a few hundred lines by file and review each in its own turn; a diff that overruns the context window gets silently truncated, and a review of half a change reads exactly like a review of all of it. Asking for one axis per pass — correctness, then style, then performance — also keeps each answer short enough to check.
For PR review, Core sizes its context window to the memory your Mac has free, rather than to a fixed 32K, which matters for long diffs. Both share the same weights, so answer quality is identical.
Large diffs are the limiting factor, not the review itself. Split anything past a few hundred lines by file and review each in its own turn; a diff that overruns the context window gets silently truncated, and a review of half a change reads exactly like a review of all of it. Asking for one axis per pass — correctness, then style, then performance — also keeps each answer short enough to check.
This guide does not claim feature parity with cloud-side workflows for “how to review a pull request without sending code to the cloud”. Specifically, the product surface in v1.11.469 covers chat, file attachment, the local agent loop, project scoping, and the model picker. Cross-device sync, team workspaces, and shared session history are out of scope and are not on the v1.9 backlog either.
Code defaults to 64K context against Core’s 32K, with both capped at 256K. The two share the same safetensors and produce identical answer quality; the difference is the chat template and decoding configuration. For PR review on diffs over 500 lines, Core’s memory-sized window keeps the whole diff plus the relevant surrounding code addressable in a single turn.
For a 500-line diff, Core’s context window holds the diff plus surrounding source. For longer diffs, paste the diff alone and reference the surrounding files by name; Outlier loads them via the project chip if needed.
4 steps, zero network requests after the model is downloaded. Asking along a single axis at a time (correctness, then style, then perf) holds quality better than a single ‘review this’ ask.
Download Outlier for MacRequires Apple Silicon (M1, M2, M3, or M4) — Intel Macs are not supported. macOS 12+.
Open weights ship constantly and most are not worth your disk. When one beats a tier Outlier already has, on hardware you already own, we say so and what it replaces. That is the only reason we email.
Your address and which site you sent it from. No IP, no user agent, no referer. Nothing is sent until you press the button. Privacy.
Outlier runs entirely on your Mac. No prompts leave the device. macOS 12+ on Apple Silicon (arm64). Apache 2.0 model weights. Back to home.