Performance and resource usage
Oasis is a Firefox-based browser with an integrated assistant. How heavy it feels depends mostly on how many tabs and extensions you run—and whether the assistant is actively working.
What this is
This page answers common questions about CPU, memory (RAM), disk, and network use. It is not a benchmark report: exact numbers vary by operating system, hardware, build, and extensions.
Is Oasis too heavy?
For everyday browsing with the assistant closed or idle, Oasis should feel close to a modern Firefox-class browser on the same machine. You are not running a large local language model in the background all day.
Resource use increases when you actively use the assistant—streaming replies, multi-step tool loops, or optional on-device embedding features in some builds. Assistant inference runs in the cloud, so much of the heavy model work is network-bound rather than pegging your GPU for local inference.
If Oasis feels slow, check what uses resources and how to keep it light before assuming the product is always "heavy."
What uses resources?
| Activity | CPU | RAM | Disk | Network | When |
|---|---|---|---|---|---|
| Normal browsing (tabs, sites) | Moderate per active tab | Grows with open tabs | Cache and profile data | Page loads | Always while browsing |
| Extensions | Varies | Varies | Minimal | Some extensions call out | When enabled |
| Assistant idle (panel open, not working) | Low | Small UI overhead | Local chat history | Minimal | Assistant visible |
| Assistant active (chat, streaming) | UI + parsing | Thread history in memory | IndexedDB chats | Higher — cloud model | While a reply is in flight |
| Multi-step agent loop | Moderate | Moderate | Minimal | Sustained while tools run | Complex tasks |
| On-device embeddings (optional) | Spikes when running | Model in RAM | Larger install | Low for local pipeline | When packaged and enabled |
| Video, games, heavy sites | High | High | Cache | Streaming media | Per site content |
Firefox baseline: tabs and processes
Oasis inherits a multi-process browser architecture: content is isolated per tab (and extensions can add their own processes). More open tabs generally means more RAM. Crashes in one tab are less likely to take down the whole window.
For deep browser tuning (hardware acceleration, cache, content process limits), see Optimize browser performance settings—a general performance reference with Mozilla documentation links.
Assistant and AI workloads
The assistant uses cloud model processing for chat (see Assistant and cloud data). While you wait for a reply:
- Network carries prompts, streaming tokens, and tool traffic.
- CPU/RAM handle the assistant UI, timeline, and local chat storage—not a full local LLM on your GPU by default.
- Busy / streaming states (Busy streaming state) indicate work in flight; avoid duplicate sends until the busy bar clears.
Long or tool-heavy tasks can keep the assistant active longer and use more of your daily token allowance—see Daily token usage bar.
Optional on-device embeddings
Some builds ship a local embedding stack (WASM + small model) for limited semantic features without calling a remote model hub. That is separate from full assistant chat:
- Disk: larger download or install footprint when embedding assets are bundled.
- RAM/CPU: used when those features run—not continuously for ordinary browsing.
Details: On-device embeddings.
How to keep Oasis feeling light
- Close tabs you are not using; each tab can hold memory.
- Review extensions—disable ones you do not need; check their impact in Task Manager.
- Use profiles to separate heavy work from everyday browsing (Manage browser profiles).
- Minimize or close the assistant when you are only reading the web, not chatting.
- Break up large agent tasks into smaller prompts instead of one endless multi-step loop.
- Restart Oasis after very long sessions if memory creeps up (common with any multi-tab browser).
Check usage yourself
Oasis includes a built-in Task Manager (same family of tool as Firefox):
- Open the application menu.
- Choose More Tools → Task Manager.
- Sort by Memory or % CPU to find heavy tabs or extensions.
Use this before blaming the assistant—often one site or extension dominates usage.
Compared to other AI browsers (qualitative)
Many AI browsers are a Chromium shell plus an always-visible chat sidebar. Oasis combines a Firefox-class browsing core with an assistant you engage when you need it. Neither approach is universally lighter; what matters is how many tabs, extensions, and active assistant sessions you run at once.
We do not publish head-to-head RAM or CPU benchmarks in this doc. Your machine and workflow matter more than a generic label of "heavy" or "light."
Notes and limits
- No fixed minimum RAM or CPU spec is listed here—requirements vary by OS and build.
- Feature flags and enterprise policies can change which optional components (embeddings, tools) ship in your build.
- Performance improvements ship over time; this page describes behavior at a high level.
Related topics
Related Documentation
Training and bonus tokens
How Oasis training works, whether anonymous training leaves the device, privacy-first trainable browser, personalized vs anonymous modes, and daily bonus tokens.
Import data from other browsers
How Oasis imports bookmarks, passwords, history, extensions, and autofill from Chrome, Edge, Safari, Brave, and more—and where that data lives relative to the cloud assistant.
Import opt-out
Users can skip browser data import during onboarding (browser.oasis.onboarding.importOptOut).
About the Author
I'm the CEO of Kahana, bringing a unique perspective from my management consulting experience at Clarkston Consulting and biomedical engineering background from Duke University. I'm focused on making the future of work more elegant through innovative technology solutions that prioritize user well-being and productivity.