Updates

Name: ARMES
Availability: InStock
Author: ARMES

New features, model additions, and platform announcements.

Introducing the Eco Plan: Serious AI, $9/mo, a Fraction of the Energy

Eco Mode proved that efficient AI doesn't mean weaker AI. Now it's a whole plan: ARMES Eco brings a dedicated agent — ARMES GPT ECO — built entirely on sparse models that activate 3–6% of their parameters per response. 10–30x more conversations per dollar, a meaningfully smaller energy footprint, and the exact same private inference

July 3, 2026

Updates7 min read

Introducing Persistent Files: A Permanent, Searchable Library for Your AI

Upload a document once and every ARMES agent can find, summarize, and read it in any chat — forever. No more re-uploading the same file into every conversation.

July 1, 2026

Updates7 min read

Introducing Projects: Context-Aware Workspaces for Your AI

Projects let you separate every body of your work in ARMES — clients, books, courses, work vs. personal — so your AI focuses on one workstream at a time, with your shared knowledge always on hand.

June 30, 2026

Updates8 min read

13 New Models. 2 New Labs. A Smarter Auto Mode on Every Tier.

ARMES now offers 48+ models from 15 AI labs. Two new labs (Alibaba's Qwen and Xiaomi's MiMo), 13 new models including Claude Opus 4.7, GPT-5.5, and Kimi K2.6, and a refreshed Auto Mode on every tier — including a brand-new Research and Synthesis routing branch that pairs Kimi K2.6 with our web search tool for low-hallucination fact-finding. No setting to flip.

April 28, 2026

Updates5 min read

Introducing Eco Mode: Powerful AI That Uses Less Energy

Eco Mode routes to highly capable Mixture-of-Experts models that activate just 3–6% of their parameters per response — consuming less energy and stretching your AI usage budget 10–30x further. Powerful AI with a smaller footprint, available with a single toggle.

March 17, 2026

Updates7 min read

A Fairer Way to Manage AI Usage: Cost-Based Budgets, Full Transparency, and Boost Packs

ARMES moves from quantity-based limits to transparent, cost-based AI usage budgets. Simple questions barely touch your budget while complex analysis uses more — reflecting reality. Plus AI Boost Packs for power users who need more, so your workflow never hits a hard stop.

March 14, 2026

Updates9 min read

Custom Instructions & Projects: How to Make Every ARMES Agent Truly Yours

Learn how to transform your ARMES agents from generic AI into personalized specialists with custom instructions and projects. Covers the two-tier instruction system, best practices backed by research on how language models process instructions, and how the ARMES Concierge can optimize your entire setup.

March 5, 2026

Updates18 min read

Every Agent in ARMES Just Got a Major Upgrade

Every agent in the ARMES library has been rebuilt with agent-to-agent collaboration, AI-managed notes, research-backed frameworks, structured methodologies, and concrete deliverables. Plus seven new agents including ARMES Concierge, Deep Research, Execution Architect, and the Future Self Designer.

March 2, 2026

Updates8 min read

February Community Update

The most significant ARMES update ever: a rebuilt ARMES GPT with agent-to-agent collaboration and AI-managed notes, six new frontier models from five AI labs, a brand new Command View workspace, platform-wide improvements, and a rewritten privacy policy. Everything you need to know.

February 20, 2026

Updates14 min read

Three Labs, Six Models, One Week: Inside the February 2026 Frontier Intelligence Update

The last two weeks have been one of the most significant periods in AI model releases we've ever seen. Three of the world's leading AI labs — Anthropic, Google DeepMind, and OpenAI — shipped major updates within days of each other. Claude Opus 4.6. Claude Sonnet 4.6. Gemini 3.1 Pro. Plus new release

February 20, 2026

Stay in the loop

New writing on private AI, delivered occasionally.

No spam. Unsubscribe anytime.RSS

Don't just read about AI. Use all of it.

Every frontier model in one private app, with intelligent routing and a memory that compounds. Start free — no setup required.

Start free

See pricing