v1.0.0 · Windows x64

Four AIs.
One answer.
Convergence

Ask Claude, GPT, Gemini, and Grok the same question. Let them debate each other across multiple rounds. Get one synthesized answer that's better than any single model could produce alone.

Runs locally · Open source · Free to use · Bring your own API keys

SIZE: 13.6 MB  ·  OS: Windows 10/11 (64-bit)  ·  VERSION: 1.0.0

One AI is a single point of failure

When you ask one AI a question, you get one perspective. That model can be confidently wrong. It can miss edge cases, hallucinate facts, fixate on one approach, or pull from biased training data.

Convergence fixes this by orchestrating a structured debate between multiple frontier models from different companies. Weak answers get challenged. Strong ones get reinforced. The final synthesized output incorporates the best ideas from all of them — and is meaningfully better than any single model's first attempt.

Built as a Software Engineering project, the entire orchestration logic lives in a Python engine, wrapped in a native Tauri v2 desktop shell. AI debating AI is the product.

convergence ~ debate.log
# Round 1 — independent answers $ claude "Use approach A" $ gpt-4 "Approach B is better" $ gemini "Approach A, but..." $ grok "Hybrid of A + B"
# Round 2 — critique & refine models converge on hybrid CONVERGENCE DETECTED
# Final synthesis output: unified answer ✓

From task to consensus in minutes

01
You submit a task "Write Python code to convert .epub files to .pdf" — anything from code to research questions.
02
Each model answers independently Round 1: every configured AI writes its first answer in isolation, no knowledge of the others.
03
Models critique each other Round 2+: every AI sees the others' answers, critiques them, refines its own response.
04
The engine detects convergence When the models start agreeing on a core answer, the debate stops automatically (safety cap: 7 rounds).
05
Final synthesis A single, unified answer combining the strongest ideas from every model.

Built to just work

Convergence RUN page
RUN — submit a task to every model
Convergence SETTINGS page
SETTINGS — local API keys, never bundled
Convergence active debate
DEBATE — watch the models converge live
Convergence history page
HISTORY — every past session, saved locally

Run it for $0

›_ The cheapest path: OpenRouter + Gemini

You can run Convergence completely free using these two providers. No credit card required. You only need one key to start, but two or more is where the magic happens.

FREE TIER

OpenRouter

One key, dozens of models. Free-tier models like Llama 3.3 70B run at $0. Sign up with Google or GitHub.

Get your key
FREE TIER

Google Gemini

Generous free-tier limits on all Gemini models. Sign in with any Google account.

Get your key
PAID

Anthropic (Claude)

Top-tier model for complex reasoning. Pay per token. Requires credit card.

Get your key
PAID

OpenAI (GPT)

The classic. Pay per token. Strong general-purpose model.

Get your key
PAID

xAI (Grok)

Elon's model. Strong on real-time and unconventional reasoning.

Get your key
$0

to start.
OpenRouter + Gemini free tiers cover everything.

Tiny footprint, fast install

Operating System
Win 10/11
Architecture
x64
Disk Space
~50 MB
Download Size
13.6 MB

No Python, Node.js, or developer tools required. Everything is bundled in the installer.

Signal from the field

convergence@telemetry: ~ $ tail -f live_feed Live
Visits Today
···
 
Total Visits
···
since deployment
.exe Downloads
···
all releases · via GitHub
Your Origin
···
 
Top origins (by session)
› collecting signal...
Visits are counted once per browser session. Download totals are pulled live from GitHub Releases (the source of truth). No cookies, no third-party trackers, no accounts.

Download Convergence now

Install in under a minute. Run debates between the smartest AIs on the planet.

Download for Windows — It's FREE
SIZE: 13.6 MB  ·  FILE: Convergence_1.0.0_x64-setup.exe