How Fast Does a GPU Pay for Itself With GIDE? Faster Than You Think
Per-token AI coding bills grow with every developer. We ran the numbers on pointing GIDE at your own GPUs, for teams that already own them and teams buying new ones.
blog
Release notes, engineering deep dives and product updates, newest first. Subscribe via RSS.
Per-token AI coding bills grow with every developer. We ran the numbers on pointing GIDE at your own GPUs, for teams that already own them and teams buying new ones.
AI coding is moving from seat plans to per-token bills. Here's why GIDE lets enterprise teams run most of that work on the Macs they already own, and keep frontier models for the hard parts.
GIDE is an AI code editor for local and cloud models. Run a model on your machine, on your own key or on your own server, behind the same approval gates.
GIDE doesn't just swap one model for another. Two models stay loaded, any model is an @ away, every model works inside the same harness, and you can plan with Claude and let a local model do the edits.
GIDE 2.4.2 finds, downloads and tests a local model that fits your machine, completes support for a model on your own server with a one-command setup, lets local models read images, and caches online conversations.
GIDE 2.4.0 keeps a second local model loaded, lets you @ mention any local or online model mid-conversation, connects to a model on your own VPS, and reads images and documents.
GIDE 2.3.6 connects any OpenAI- or Anthropic-compatible endpoint with your own key, makes Ornith 1.5 9B the default local model, answers questions about GIDE itself, and asks before a tool reaches outside your project.
GIDE 2.3.5 brings the full agent to the terminal with the GIDE CLI, ships Linux builds for the app and the CLI, adds Folder Trust before any read or write, and opens GIDE into a conversation.
GIDE 2.3.3 reworks the first-run experience: a Welcome card with three direct actions, model downloads inside the chat welcome view, and a model picker that shows every state a model can be in.
GIDE 2.3.0 moves model weights out of the app so updates never re-download them, manages Ornith downloads in the app with resume and verification, and rebuilds the chat around compact tool chips and live thinking.
GIDE 2.2.0 auto-configures more local model families, adds online models on your own API key, introduces Skill Files and MCP servers, and gives every model temperature and GPU offload controls.
GIDE is an AI code editor that runs local models on your machine, online models on your own key, or models on your own server. Here's what it is, how it works, and who it's for.
GIDE 2.1.0 introduces Individual, Pro and Enterprise subscription tiers with server-side feature access, moves command approval into the chat transcript, and adds a collapsible Thinking block with a live timer.
GIDE 2.0.0 adds Private Cloud mode, puts every file write and shell command behind your approval with the Write Gate and Command Gate, gives the agent a dedicated Git tool, and rewrites GIDE Wiki's architecture graph.
GIDE 1.106.0 rebuilds the agent around a plan-gate architecture, adds Context Memory, a refactor engine for 21 languages, 30+ new agent tools and a redesigned code indexer.
GIDE 1.105.4 brings GIDE to Linux, packaged as AppImage, .deb and .rpm, and validated on Ubuntu 24.04 under WSL2.
GIDE 1.105.3 makes Ornith 1.0 9B the default bundled model on every platform, adds downloadable Qwen, Gemma and DeepSeek models plus any GGUF, and fixes a batch of UI issues.
No posts with this tag yet.