Compare AI models side by side, on the same context.

Three steps

1. Branch once per model. In a chat's message box, press ⌘B (Ctrl+B on Windows and Linux) once for each model you want to try.
2. Pick a model in each branch. Choose from the model dropdown in each new chat.
3. Send the same prompt. The answers sit side by side on the canvas.

• Step-by-step in the docs

A fair test, because the context is the same

Branches that share a parent answer from the same context.

• Same starting point. Each branch reads the parent's whole conversation, plus every source connected to the parent.
• Same setup. A new branch copies the parent's mode, persona, reasoning level and connectors, so the model is the only thing you change.
• No peeking. Sibling branches can't see each other, so one model's answer never shapes another's.

Compare on your own sources

Most real questions come with material. Connect a PDF, a web page or a YouTube video to the parent chat, and every branch reads it. Nothing is summarized or trimmed: each connected source is sent in full with every message, so you see how each model handles the same evidence.

One thing to know: enough big sources can exceed a smaller model's context window. Disconnect what the chat doesn't need, or pick a model with a larger window.

• How connections work

Any model, from wherever you run it

Mix models from different places in one comparison.

• Your Claude or ChatGPT plan. Slashspace drives the official Claude Code and Codex CLIs on your computer, signed in with your own account. Usage counts against your plan.
• Your API keys. OpenAI, Anthropic, Google, Perplexity, xAI, OpenRouter, DeepSeek, Groq and more, billed by each provider.
• Local models. Ollama runs open models on your computer, with no per-message cost. LM Studio and vLLM connect as custom providers.
• Optional built-in AI. Slashspace's hosted models, with no setup.

• Every way to run AI in Slashspace

Bring the answers together

Reading four answers is only half the job. Connect the finished branches into a new chat, and it reads every one of them in full. Ask it where the models agree, where they differ, and what to trust, without copy and paste.

Keep comparisons readable

• Name each branch after its model, and color them.
• Tidy Up lays the canvas out left to right, each branch next to its parent.
• Group a finished comparison into one labeled box.
• Collapse branches to their last exchange when you're done.

• More on branching

Tips for a clean comparison

• Reasoning effort (Low to Max) reaches models on your own OpenAI, Anthropic or Google keys. Other models ignore it, so keep that in mind when you compare.
• Claude Code and Codex are agents that work in a folder you choose. Pick a workspace folder before the first message.
• Isolated context carries over. If the parent chat uses Isolated context, its branches ignore their incoming connections too. Switch the parent to Summarized context before you branch.

Other ways to compare models

Comparing models isn't unique to Slashspace. TypingMind has multi-model responses, and Msty has Split Chats, and both are good at it. What Slashspace adds is the canvas: every model answers from the same wired-in sources, the comparison stays on the map next to the rest of your research, and you can connect the answers into one synthesis chat.

• TypingMind comparison
• Msty comparison
• More branching apps, compared

Slashspace is a desktop app for Mac, Windows and Linux. Web and mobile apps are coming soon.