AI WAR ROOM
Developer tools
Every AI model on one board, ranked by real votes.
A free web leaderboard that RVGE has run since spring 2026 to follow AI models as they come out. Over 400 language models are ranked by arena ELO from human head-to-head votes, with price, context window, speed and benchmark scores beside each one, plus boards for coding agents, image, video and voice. Every number is pulled from a named source on a schedule; nothing is typed in by hand.
Live web app, no account needed · Free, nothing to buy
- Status
- Public beta Live · free · aiwarroom.app
- Platform
- Web
- Last work
- Built with
- React 18 · Vite · Netlify Functions · Netlify Blobs · SVG charts
- For
- Developers and teams choosing which AI model to build on, and anyone following the race.
What AI WAR ROOM does
Ranked by real votes
Language models are ordered by arena ELO from human head-to-head votes on arena.ai, shown with confidence intervals and vote counts, and refreshed every six hours.
Price and speed beside the score
Each model carries input and output pricing and context window from OpenRouter, and intelligence index, speed and latency from Artificial Analysis, so the charts show value, not only rank.
Code Ops
Coding agents on SWE-bench, with the harness and the model kept apart: tasks resolved, cost per task and calls per task.
Image, video and voice boards
Image and video arenas from crowd votes, and speech recognition from the Hugging Face Open ASR leaderboard, with word error rate and speed.
Provider War
The same model across inference hosts: price, context, quantisation and uptime per endpoint from OpenRouter, with the cheapest and the most reliable host called out.
Battle, Race and Planner
Put two models head to head, race them on tokens per second, or tell the Mission Planner what you need and get a scored recommendation.
History that remembers
A scheduled job stores a snapshot whenever the numbers move, so rank changes over one, seven and thirty days and the trend charts on model pages come from real history.
Never an empty page
Every source is cached on its own. When a feed fails, the last valid copy is shown with its age, and a missing figure reads N/A instead of an estimate.
Where your data goes, and what to expect.
Free to use, no account. The site shows Google AdSense ads and uses Google Analytics, which no RVGE desktop product does. Its own privacy page lists both.
Every score is third-party data: arena.ai, OpenRouter, Artificial Analysis, SWE-bench and the Hugging Face Open ASR leaderboard, each named where its numbers appear. AI WAR ROOM runs no benchmarks of its own and is not affiliated with any lab or data provider.
Rankings move. A model that leads today can be fifth next week, and the site says so in its own guide.
What is still missing.
This list is the work between today's build and a version anyone can rely on. It gets shorter, it is never hidden.
- Consent before analytics. Google Analytics starts when a page loads; a consent step in front of it, and of the ads, is the next job.
- An About page that names RVGE. Today the site says only that it is a small independent project.
- Alerting when a feed goes stale. The board shows the age of its cached copy, but nobody is paged when a source keeps failing.
Recent changes.
- Site
AI WAR ROOM joins the catalogue
The free AI model leaderboard RVGE has run since spring 2026 is now listed as a product, with its data sources, its test count and its open issues written down on its page.
Source · rvge.dev
- Feature
Metered feed, capped
The Artificial Analysis Data API is called at most twice a day, and a failed call is not retried for twelve hours, so the site stays inside the feed's limits and keeps serving the last good copy meanwhile.
Source · AI-WAR git e292e11
- Feature
Rankings as charts, never an empty page
Rankings are drawn as bar charts with a model selector, a benchmark grid and openness filters. When an upstream fails, the page falls back to the published leaderboard pages or the cached copy; a zero is never shown as a measurement.
Source · AI-WAR git 4965963, 9b20386, 8564e13
- Milestone
From leaderboard to command center
Code Ops, image and video arenas, Voice Comms, Provider War, Benchmarks, Battle Mode, Speed Race and the Mission Planner join the LLM leaderboard, all reading one normalised dataset served by a single function.
Source · AI-WAR git 90094fc, dc56950
Other RVGE products
RVGEBRAIN
One workbench for every coding agent you already use.
A Windows development environment built around AI coding agents. Run Claude Code, Codex and other CLI agents side by side, see at a glance which one needs you, review their edits in a real editor, and keep every session alive across restarts.
Explore
RvgeShot
Screenshots and recordings that never leave your PC.
Screen capture and recording for Windows. Press Print Screen and it captures the screen exactly like the key always has, then crop, annotate, record and keep it all in a local gallery on your own machine.
Product page
SpkFlw
Hold a key, speak, and the text lands where you type.
Push-to-talk dictation for Windows. Hold Ctrl and Win, speak, let go, and the transcript is pasted into whatever field you were already typing in. Speech recognition runs on a provider you connect with your own key.
Product pageTry AI WAR ROOM.
It is free and runs in any browser. Tell us what is missing or wrong, it decides what gets built next.