Ask two AIs the same question. Keep the one that's actually right.
For developers and researchers who pay for GPT, Claude, and Gemini — and are tired of guessing which one to trust. Run them side by side, vote on the winner, and stop wasting money on the wrong model.
Free · no credit card · 50 free credits when you sign up · bring your own API keys →
Stop guessing which AI to trust.
Real answers from two live models on our keys — free to try, no account. Sign up to run unlimited prompts across 20+ models with your own keys.
Truth Arena
Compare models side-by-side, expand the Truth panel, re-ask on another lab in one click.
Every lab first-class
Sakana Fugu, Perplexity Sonar (live/deep research), Grok, GPT, Claude, Groq, Gemini, OpenRouter — each lab treated as first-class.
BYOK by default
You pay labs for tokens. You pay Verum for the workbench. Clear and simple.
Memory that sticks
History, profile memory, and documents that inject when relevant.
Pro console density
Right rail, presets, capability cards, and View Code — a calm multi-lab workspace.
Private multi-tenant
Encrypted keys, server-side proxy, row-level isolation. No ads in the chat. Ever.
Frequently asked questions
What Verum is, which models it compares, and how bring-your-own-key (BYOK) works.
What is Verum?
+
Verum is a multi-model AI workbench. You bring your own API keys (BYOK) for the labs you use — Sakana Fugu, OpenAI GPT, Anthropic Claude, Google Gemini, xAI Grok, Moonshot Kimi, Perplexity Sonar, Groq and OpenRouter — and run, compare and vote on their answers side by side in one private console.
Which AI models can I compare in Verum?
+
Verum supports GPT (OpenAI), Claude (Anthropic), Gemini (Google), Grok (xAI), Kimi K3/K2 (Moonshot), Sakana Fugu, Perplexity Sonar for live web and deep research, Llama and GPT-OSS on Groq, plus hundreds more through OpenRouter and any OpenAI-compatible custom endpoint.
What does BYOK (bring your own key) mean?
+
You create API keys at each lab and paste them into Verum's encrypted settings. The labs bill you for tokens; Verum charges only for the workbench. Your keys are encrypted at rest and used solely to run your own requests.
How is Verum different from ChatGPT, Poe or OpenRouter?
+
ChatGPT gives you one lab. Poe and OpenRouter route to many models but hide them behind their own billing. Verum keeps every lab first-class, uses your own keys, and adds a blind Truth Arena so you can objectively compare two models on the same prompt and keep what actually works.
Can I use Kimi (Moonshot AI) in Verum?
+
Yes. Add your Moonshot / Kimi API key in Settings under the Moonshot AI (Kimi) provider. Verum talks to Kimi's OpenAI-compatible endpoint at https://api.moonshot.ai/v1, then pick Kimi K3 or Kimi K2.5 in the model menu.
What is the Truth Arena?
+
The Truth Arena runs the same question on two different labs at once. You read both answers, expand the Truth panel, and vote for the better one. Over time Verum builds your personal Best pick from those blind votes.
How much does Verum cost?
+
The workbench is $12/month or $120/year and includes memory, side-by-side compare, export and higher limits, with a 14-day Pro trial and no ads. Model tokens are billed by each lab on your own keys (BYOK).
Is my data and API key private?
+
Yes. Keys are encrypted at rest, requests are proxied server-side, tenants are isolated with row-level scoping, and there are no ads in the chat — ever.
Every model. Your keys. Your truth.
Truth Arena runs the same question across labs. BYOK — you pay models, we sell the cockpit. No ads.