SiteList review score
Exceptional technical integrity and benchmark-driven content. Undermined by a 30% internal link failure rate and unoptimized asset payloads.
Reviewed on Aug 22, 2026
Local — Zero (!) friction local AI for your Mac
basecompute.co
Unclaimed listingWhat it does
Local is a macOS application that runs AI inference entirely on-device without cloud dependency or per-token fees. It automatically optimizes performance for Apple Silicon via BaseRT, supports on-machine chat, coding, and meeting transcription, and allows users to connect external API keys for frontier models when needed.
Two ratings
Verified user rating
—
No verified reviews yet — the evidence locker is empty.
Screenshots
Captured during our latest review on Aug 24, 2026
Pricing details
Zero per-token fees; uses open local models with the option to connect user-provided API keys for frontier model access. Billed monthly.
| Plan | Price | Details |
|---|---|---|
| Local (Base) | $0 monthly per app | Runs open local models on-device with zero per-token fees. Requires macOS · Uses user-provided API keys for optional frontier models View source |
Key features
Chat with files
chat integration
Ask questions, draft writing, and summarize attached PDFs directly within the app.
Code in folder
coding integration
Point the app at a project directory to read, edit, and run code locally without uploading the codebase.
Meetings transcription
meetings integration
Record and transcribe meetings on-device with automatic speaker labeling.
Memory control
memory management
Maintain a visible, editable list of lasting facts about the user; sensitive items are restricted from leaving the Mac.
Office Mode
office integration
Run a large model on a capable machine and access it remotely from Local apps on other laptops in the same workspace.
Who it’s for
Chat with files
Ask questions, draft writing, and summarize attached PDFs directly within the app.
Code in folder
Point the app at a project directory to read, edit, and run code locally without uploading the codebase.
Meetings transcription
Record and transcribe meetings on-device with automatic speaker labeling.
Integrations
DeepSeek
Gemma
Llama
Mistral
Phi
Qwen
Whisper
Buyer questions
Questions buyers actually ask
Is Local by Base Compute faster than llama.cpp?
Base Compute claims performance up to 6.4x faster than llama.cpp for prefill operations. These claims are anchored in a formal academic paper (arXiv:2607.00501), providing a high level of technical integrity compared to typical marketing claims.
What hardware does Local support?
The product is specifically optimized for AI engineers using Apple Silicon and high-end GPUs. The messaging and technical documentation are tuned for power developers who prioritize on-device performance and privacy.
Why is the site's performance score low?
The site received a 48/100 for performance due to a 3.5MB total payload. This bloat is primarily caused by unoptimized PNG screenshots and render-blocking resources that delay content visibility for users.
Is Local suitable for enterprise use?
While the technical foundation is strong, the site currently lacks social proof and enterprise-specific case studies. It is best suited for individual AI engineers and privacy-focused organizations at this stage.
Answers are based on SiteList's latest evidence review. Read the full review.
The expert review
Local by Base Compute earns a 78/100, distinguished by its exceptional technical integrity and benchmark-driven content. However, the platform's digital presence is currently undermined by significant infrastructure issues, including a 30% internal link failure rate and unoptimized asset payloads.
Technical substance is high, but infrastructure rot persists.
Verified user reviews
0 verifiedNo reviews yet. The evidence locker is empty.
Reviews from verified users appear here once approved.
Reviews are from the SiteList community. Some reviewers received incentives — disclosed in the review, always. The verdict is never for sale.
Scores across 13 dimensions
Brand & experience
Performance
Content & editorial
Trust & decision surfaces
Search & discoverability
Docs & measurement
- Domain
- basecompute.co
- Category
- Infrastructure & hosting
- Pricing
- Free · $0/mo
- Platforms
- desktop
- Last checked
- Aug 22, 2026
- Public dimensions
- 13
Similar tools
Home
CodeBurn — See where your AI coding spend actually goes screenshot 1 of 4: Home
CodeBurn — See where your AI coding spend actually goes
Score 84 out of 100 — Strongcodeburn.app
CodeBurn is a free, open-source, local-first tool that tracks AI coding token usage and cost across dozens of developer tools. It breaks down spend by model, project, and task using session files already on the user's disk, with no data uploaded.
Whop CLI — Run your entire business from the terminal
Score 82 out of 100 — Strongwhop.sh
The Whop CLI gives developers, creators, and automation agents command-line access to Whop API resources.
Home
Shieldstral — Define safety at runtime for text and images screenshot 1 of 4: Home
Shieldstral — Define safety at runtime for text and images
Score 81 out of 100 — Strongmistral.ai
Shieldstral introduces a 3B open-weights multimodal safety classifier that outperforms models up to 7x its size.
Home
Soup CLI — Fine-tune an 8B LLM on a 4 GB laptop GPU screenshot 1 of 4: Home
Soup CLI — Fine-tune an 8B LLM on a 4 GB laptop GPU
Score 81 out of 100 — Strongtrysoup.dev
Soup CLI is an open-source command-line toolkit for LLM post-training and fine-tuning on local hardware. It uses layer streaming to fit 8B models into 4 GB VRAM, auto-generates training configs, gates saves via evals, and supports migration from other fine-tuning tools.