What frontier AI models does ARMES include?

Name: ARMES
Availability: InStock
Author: ARMES

ARMES gives you private access to 21+ frontier AI labs: ChatGPT (GPT-5.2, GPT-5, GPT-4.1), Claude (Opus 4.8/4.7/4.6/4.5, Sonnet 4.6/4.5, Sonnet 3.7:Thinking, Haiku 4.5), Gemini (3.1 Pro, 3.0 Pro, 2.5 Pro, 3.0 Flash, 2.5 Flash, 3.1 Flash Lite), Grok, DeepSeek (V3.2, R1), Perplexity (Sonar Pro, Deep Research), Meta (Llama 4 Maverick), Inception (Mercury 2), ByteDance (Seed 2.0), Deep Cogito (Cogito v2.1), NVIDIA (Nemotron), and more. No ads. No data retention. No training on your conversations. Auto Mode routes to the best AI for each task. ARMES Deep Research powered by Perplexity delivers cited, multi-source investigations. One subscription replaces 5+ separate AI subscriptions ($20–110+/mo) — $19/month launch price (regular $25/month). New frontier models within 24 hours of release.

AI Models & Labs

Every top model. One app.

ChatGPT, Claude, Gemini, Grok, DeepSeek, Perplexity and more — in one private interface that picks the best model for every task, with new releases added within a day.

Start free

21+ labs · Private inference by default · Privacy Policy

21+ labs

All top providers

Tools built in

Research, search, code

Private inference

Nothing stored

Auto mode

Best model, automatically

Private AI labs

Every frontier lab, in one place.

Each provider below processes your request and forgets it — nothing is stored, profiled, or monetized, and the labs never see who you are.

Plans are cumulative — each plan includes everything below it.Free= every planEco= Eco, Pro & UltraPro= Pro & UltraUltra= Ultra

Private

The Specialists

OpenAI

Hard math, advanced science, and security.

Private

The Poets & Coders

Anthropic

Nuanced writing, heavy coding, strategy & empathy.

11models

Opus 5.5Ultra Opus 5Ultra Opus 4.8Ultra Opus 4.7Ultra Opus 4.6Ultra Opus 4.5Ultra Sonnet 5.5Pro Sonnet 5Pro Sonnet 4.6Pro Sonnet 4.5Pro Haiku 4.5Pro

Explore Anthropic dossier

Private

The Analysts

Google

Large-context analysis and data processing.

14models

Gemini 3.1 Flash TTSPro Gemini 3.1 ProPro Gemini 2.5 ProPro Gemini 3.8 FlashEco Gemini 3.7 FlashEco Gemini 3.6 FlashEco Gemini 3.5 FlashEco Gemini 3.0 FlashEco Gemini 2.5 FlashEco Gemini 3.5 Flash-LiteEco Gemini 3.1 Flash-LiteFree Nano Banana 2 LitePro Nano Banana 2 FlashPro Nano Banana 2 ProPro

Explore Google dossier

Private

The Underdogs

xAI

𝕏 integration, real-time sentiment/trend parsing.

5models

Grok 4.7Pro Grok 4.6Pro Grok 4.5Pro Grok 4.3Pro Grok 4.20Pro

Explore xAI dossier

Private

The Logicians

DeepSeek

Versatile reasoning at the lowest cost.

7models

V4 Pro 0813Eco V4 ProEco V4.1 FlashFree V4 Flash-0731Free V4 FlashFree R1Eco V3.2Free

Explore DeepSeek dossier

Private

The Researchers

Perplexity

Cited research and deep investigation.

3models

sonar-pro-searchPro sonar-deep-researchUltra sonar-proPro

Explore Perplexity dossier

Private

The Polyglots

Alibaba

Multilingual mastery and agentic coding.

2models

Qwen 3.8-2.4t-a95bPro Qwen-3.8-27bPro

Explore Alibaba dossier

Private

The Omnimodals

Xiaomi

Omnimodal perception and long-horizon agents.

4models

MiMo-V2.6-ProEco MiMo V2.5 ProEco MiMo-V2.6-FlashFree MiMo V2.5Free

Explore Xiaomi dossier

Private

The Coordinators

Moonshot AI

Multi-agent reasoning, math, deep research.

5models

Kimi K3Pro Kimi-k2.7-codeEco Kimi-k2.6Eco Kimi-k2.5Eco Kimi-k2Eco

Explore Moonshot AI dossier

Private

The Architects

Z.ai

Agentic coding, low hallucination, planning.

6models

GLM-5.3 FlashFree GLM-5.3Eco GLM-5.2Eco GLM-5.1Eco GLM-5Eco GLM-4.7Eco

Explore Z.ai dossier

Private

The Operators

MiniMax

Autonomous coding and agentic tasks.

4models

minimax-m3Free minimax-m2.7Free minimax-m2.5Free minimax-m2.1Free

Explore MiniMax dossier

Private

The Pragmatists

Tencent

Grounded answers, steady agents, honest limits.

2models

Hy4Eco Hy3Free

Explore Tencent dossier

Private

The Democratizers

The efficient pool.

Models marked Eco use sparse architectures that activate 3–6% of their parameters per response — a fraction of the compute, cost, and energy. They power the Eco plan and Eco Mode on every Pro & Ultra agent.

Intelligence per watt

Every model runs zero-data-retention.

We only route to endpoints that don't retain or train on your data. If a provider can't offer that, it isn't on ARMES.

How it works

Frontier Auto Mode

The right engine, automatically.

ARMES analyzes prompt complexity, domain specialties, latency budgets, and context limits in milliseconds—routing to the optimal engine across 21+ private labs without compromise.

Router Telemetry Active•Zero Data Retention Pipeline

Classification Latency: <14ms

Input Query Signal

“Prove this FrontierMath Tier 4 problem regarding modular forms, then audit the disassembled binary routine for memory-safety vulnerabilities.”

Detected Dimensions

Domain:Formal Math & Cryptography

Verification:Deterministic Proof

Compute Tier:Recurrent Depth Reasoning

Dispatched

OpenAI

GPT-6 Astra

Apex Formal Reasoning Flagship

MathArena 90.7% & SRE-Bench 88.0% SOTA. Routed to Astra for recurrent depth reasoning where mathematical proof, binary decompilation, and zero-day containment are paramount.

MathArena 90.7%FrontierMath 97.6%SRE-Bench 88.0%

Explore GPT-6 Astra specifications in Docs

Read the technical architecture behind Auto Mode routing in documentation

Best model per task

Different work, different strengths.

No single model wins at everything. Here's where each lab shines — and with Auto mode, you don't have to remember any of it.

Interactive coding & systems

In-editor full-stack development, database architecture, and complex engineering with industry-leading execution.

Grok 4.6

xAI

Dossier

Literary prose & candid review

High-stakes writing with genuine human cadence, empathetic counsel, and rigorous flaw disclosure.

Claude Opus

Anthropic

Dossier

Hard math & binary security

Deterministic proofs, advanced science, decompilation, and security containment where verifiability is everything.

GPT-6 Astra

OpenAI

Dossier

1M documents & fast data

Million-token context for whole codebases and filings, paired with sub-second biomedical and chart reasoning.

Gemini 3.8 Flash

Google

Dossier

Cited web research

Current, source-backed answers from the live web with cited domain references.

Sonar

Perplexity

Dossier

High-efficiency everyday tasks

Mixture-of-Experts routing for routine notes, brainstorming, and daily questions at 10–30× lower compute cost.

Eco MoE Pool

Across labs

Dossier

Wield all of them. Privately.

One app, every top model, private by default. Show up and get to work.

Start free

How privacy works

AI Models & Labs

Every top model. One app.

ChatGPT, Claude, Gemini, Grok, DeepSeek, Perplexity and more — in one private interface that picks the best model for every task, with new releases added within a day.

Start free

21+ labs · Private inference by default · Privacy Policy

21+ labs

All top providers

Tools built in

Research, search, code

Private inference

Nothing stored

Auto mode

Best model, automatically

Private AI labs

Every frontier lab, in one place.

Each provider below processes your request and forgets it — nothing is stored, profiled, or monetized, and the labs never see who you are.

Plans are cumulative — each plan includes everything below it.Free= every planEco= Eco, Pro & UltraPro= Pro & UltraUltra= Ultra

Private

The Specialists

OpenAI

Hard math, advanced science, and security.

Private

The Poets & Coders

Anthropic

Nuanced writing, heavy coding, strategy & empathy.

11models

Opus 5.5Ultra Opus 5Ultra Opus 4.8Ultra Opus 4.7Ultra Opus 4.6Ultra Opus 4.5Ultra Sonnet 5.5Pro Sonnet 5Pro Sonnet 4.6Pro Sonnet 4.5Pro Haiku 4.5Pro

Explore Anthropic dossier

Private

The Analysts

Google

Large-context analysis and data processing.

14models

Explore Google dossier

Private

The Underdogs

xAI

𝕏 integration, real-time sentiment/trend parsing.

5models

Grok 4.7Pro Grok 4.6Pro Grok 4.5Pro Grok 4.3Pro Grok 4.20Pro

Explore xAI dossier

Private

The Logicians

DeepSeek

Versatile reasoning at the lowest cost.

7models

V4 Pro 0813Eco V4 ProEco V4.1 FlashFree V4 Flash-0731Free V4 FlashFree R1Eco V3.2Free

Explore DeepSeek dossier

Private

The Researchers

Perplexity

Cited research and deep investigation.

3models

sonar-pro-searchPro sonar-deep-researchUltra sonar-proPro

Explore Perplexity dossier

Private

The Polyglots

Alibaba

Multilingual mastery and agentic coding.

2models

Qwen 3.8-2.4t-a95bPro Qwen-3.8-27bPro

Explore Alibaba dossier

Private

The Omnimodals

Xiaomi

Omnimodal perception and long-horizon agents.

4models

MiMo-V2.6-ProEco MiMo V2.5 ProEco MiMo-V2.6-FlashFree MiMo V2.5Free

Explore Xiaomi dossier

Private

The Coordinators

Moonshot AI

Multi-agent reasoning, math, deep research.

5models

Kimi K3Pro Kimi-k2.7-codeEco Kimi-k2.6Eco Kimi-k2.5Eco Kimi-k2Eco

Explore Moonshot AI dossier

Private

The Architects

Z.ai

Agentic coding, low hallucination, planning.

6models

GLM-5.3 FlashFree GLM-5.3Eco GLM-5.2Eco GLM-5.1Eco GLM-5Eco GLM-4.7Eco

Explore Z.ai dossier

Private

The Operators

MiniMax

Autonomous coding and agentic tasks.

4models

minimax-m3Free minimax-m2.7Free minimax-m2.5Free minimax-m2.1Free

Explore MiniMax dossier

Private

The Pragmatists

Tencent

Grounded answers, steady agents, honest limits.

2models

Hy4Eco Hy3Free

Explore Tencent dossier

Private

The Democratizers

The efficient pool.

Intelligence per watt

Every model runs zero-data-retention.

We only route to endpoints that don't retain or train on your data. If a provider can't offer that, it isn't on ARMES.

How it works

Frontier Auto Mode

The right engine, automatically.

ARMES analyzes prompt complexity, domain specialties, latency budgets, and context limits in milliseconds—routing to the optimal engine across 21+ private labs without compromise.

Router Telemetry Active•Zero Data Retention Pipeline

Classification Latency: <14ms

Input Query Signal

“Prove this FrontierMath Tier 4 problem regarding modular forms, then audit the disassembled binary routine for memory-safety vulnerabilities.”

Detected Dimensions

Domain:Formal Math & Cryptography

Verification:Deterministic Proof

Compute Tier:Recurrent Depth Reasoning

Dispatched

OpenAI

GPT-6 Astra

Apex Formal Reasoning Flagship

MathArena 90.7% & SRE-Bench 88.0% SOTA. Routed to Astra for recurrent depth reasoning where mathematical proof, binary decompilation, and zero-day containment are paramount.

MathArena 90.7%FrontierMath 97.6%SRE-Bench 88.0%

Explore GPT-6 Astra specifications in Docs

Read the technical architecture behind Auto Mode routing in documentation

Best model per task