Eco Mode: High-Efficiency Intelligence
How Eco Mode leverages Mixture-of-Experts architectures to deliver 95%+ frontier intelligence at 10-30x lower energy and inference cost.
The Philosophy of Eco Mode
Frontier dense models (running hundreds of billions or trillions of parameters simultaneously for simple queries) consume immense electrical power, water cooling, and inference capital. Yet the vast majority of everyday cognitive tasks—summarizing emails, structuring bullet points, checking grammar, or drafting basic functions—do not require a 2+ trillion-parameter supercomputer.
Eco Mode is an architectural breakthrough in sustainable AI. It routes prompts through high-performance Mixture-of-Experts (MoE) and distilled open-weight architectures that activate only the specialized subnetworks needed for the task.
Unified Mode Selector: How the Eco Plan & Eco Mode Align
Every agent on ARMES displays Auto (🧠), Max (🔥), and Eco (🌿) in its top-right Mode Selector:
- The Eco Subscription ($9/month —
ARMES GPT ECO): On the Eco plan, Auto Mode (Eco Auto) and Eco Mode (🌿) use the exact same high-efficiency routing logic across the sparse Mixture-of-Experts (MoE) pool, while Max Mode (Eco Max) locks routing to the highest-capability reasoning models in the Eco pool (and manual selection lets you pin any model from the Eco + Free pool). Having Auto, Max, and Eco onARMES GPT ECOgives you a unified Mode Selector across every agent and makes upgrading or transitioning plans effortless. - The Versatile Eco Mode Toggle (
🌿on Pro & Ultra): Already on Pro (ARMES GPT PRO) or Ultra (ARMES GPT ULTRA)? Every agent includes Eco Mode right besideAutoandMax. UseAutoorMaxfor work that demands your tier's frontier pool; toggleEco Modefor drafts, brainstorms, code reviews, and daily conversations—stretching your daily and monthly compute allowance 10–30× further. (On the Free tier,AutoandMaxare active across the Free pool, whileEco Modeis visible in the Mode Selector with an upgrade CTA.)
Intelligence per Watt
The Eco Model Pool
The Eco routing pool comprises high-efficiency, specialized Mixture-of-Experts architectures carefully selected for exceptional intelligence per watt.
You can view the models active across all tiers on the AI Access Portal.
When to Choose Eco, Auto, or Max
- Choose Eco Mode (
🌿): For brainstorming, drafting notes, editing prose, quick calculations, translating text, formatting JSON/tables, and all routine daily questions where minimal energy and compute footprint are preferred. - Choose Auto Mode (
🧠): For everyday general work where you want adaptive routing that right-sizes compute to the prompt—automatically escalating to frontier power when needed. - Choose Max Mode (
🔥): For complex architectural system design, formal mathematics, whole-repository engineering, or multi-step reasoning where uncapped apex frontier performance is essential. See Routing Architecture.