Stop trusting
one AI.
The Drux Consensus API fans your prompt out to multiple frontier models simultaneously, scores how much they agree, and returns a single synthesised answer with a confidence score. One call. Real signal.
3 models · simultaneously
Remote work generally improves productivity for individual contributors. Studies show 13-20% output gains when distractions are managed. The key variables are home setup quality and task type.
The productivity question depends heavily on role. Engineers and writers thrive remotely. Sales, mentorship, and creative brainstorming benefit from physical presence. A hybrid model captures both.
Evidence is genuinely mixed. Remote work increases autonomy and reduces commute stress, which correlates with output. But collaboration overhead rises and junior employees develop skills more slowly without osmotic learning.
Consensus output
Leans one way
Consensus score / 100
Synthesis
Most models agree remote work can boost individual focus and deep work, but the evidence is mixed for collaborative tasks. The consensus leans positive for knowledge workers with structured routines, while flagging that team cohesion and onboarding suffer without intentional in-person time.
Preview — click Run to try your own query
How it works
01
One API call
Send your prompt once. We fan it out to all requested models in parallel — no serial waits, no rate-limit juggling.
02
Parallel execution
All models run simultaneously. Results come back in seconds, not minutes. We handle retries, timeouts, and fallbacks.
03
Consensus score
We embed all responses, cluster by semantic similarity, and compute Shannon entropy. The result: a 0–100 confidence score and a single synthesised answer.
Integrate in minutes
curl -X POST https://drux.space/api/v1/consensus \
-H "Authorization: Bearer drx_your_key" \
-H "Content-Type: application/json" \
-d '{
"query": "Is remote work better for productivity?",
"models": [
"meta-llama/llama-4-maverick",
"deepseek/deepseek-chat",
"google/gemma-3-27b-it"
]
}'Built for real use cases
🔬
Research & fact-checking
Cross-verify claims across multiple frontier models. Low consensus flags contested or uncertain information before you publish.
🏗️
Product decisions
Run technical or strategic questions through multiple models and score how much they agree. High consensus = strong signal.
📊
Content moderation
Multi-model scoring dramatically reduces false positives. A single model's blind spots get corrected by the group.
🤖
Agent pipelines
Use consensus as a confidence gate. Only proceed when models agree above a threshold. Reduce hallucinations in production.
💬
Customer support
Route ambiguous tickets to human review when AI consensus is low. Automate only the high-confidence answers.
📝
Document review
Surface sections where models disagree on interpretation — contracts, policies, medical texts. Disagreement = review flag.
Reading the score
0 – 25
Highly contested
Models fundamentally disagree. Treat this topic as unresolved. Human review recommended.
26 – 50
Divided
Meaningful disagreement. Multiple valid perspectives exist. Useful to surface the splits.
51 – 75
Leans one way
General agreement with notable exceptions. The synthesis is reliable but worth reading.
76 – 100
Converging
Strong consensus across models. High confidence in the synthesised answer.
Simple credit pricing
1 credit per model
A 3-model consensus call costs 3 credits. A 5-model call costs 5. No subscriptions, no minimums. Credits never expire.
400
credits for $5
1,000
credits for $10
3,000
credits for $30
Ready to build with consensus?
Get your API key in 30 seconds. Free tier included.