Which models can I use?

60+ frontier models including GPT-4o, Claude Opus/Sonnet/Haiku, Gemini 2.5 Pro, Llama 3.3, Mistral Large. Pick from the rail or let the router decide.

What's the context window?

Up to 1M tokens on Gemini 2.5 Pro and Claude 1M-context tier. Other models use their native ceiling; the UI reflects the active model's limit.

Which file types can I upload?

PDF, DOCX, TXT, and CSV. Files are embedded and become first-class chat context, with no separate 'doc mode' to switch into.

How does Web chat work?

Web chat augments answers with live web-search results from a dedicated surface (AIWebChat extension). Use it when freshness matters.

Does Memory cross conversations?

Yes. Save instructions once (tone, role, preferences) and every future chat inherits them via our long-term-memory store. Clear or edit any time.

Can I organize chats?

Folders let you group conversations by project. Move chats between folders, pin important threads, rename freely.

Yes. Every surface exposes a server-to-server JSON endpoint. Vision and file upload accept multipart payloads.

Are my chats private?

Conversations belong to your workspace. We don't train models on your data; vendor providers operate per their published policies.

Chat Studio

One chat. Every frontier model. Files, the web, vision, memory.

Switch between GPT-4o, Claude, Gemini, Llama and Mistral mid-thread. Drop in PDFs, search the live web, paste an image, or save long-term instructions that follow you across conversations. One surface, five production workloads, backed by the full AresGen model catalog.

Start Free View Pricing

60+ models · 1M-token context · Multimodal

60+ frontier models routed at send-time: OpenAI, Anthropic, Google, Meta, Mistral
Drop PDFs, DOCX, TXT or CSV; the file becomes chat context
Live web search augmentation on demand (Web chat surface)
Memory + folders: instructions and organization persist across sessions

Refactor this Laravel job into a queued service.

Sure. I'll extract the dispatch logic into a ChatJobService and wire it through ShouldQueue. Here's the diff…

Now compare it against Claude Opus.

GPT-5GPT-5 CodexGPT-4oGPT-4.1Claude Opus 4.7

60+ frontier models4 file types ingestedLive web searchVision + memory

By the numbers

One subscription. Five surfaces. The full frontier-model catalog.

Models: 60+; Routed at send-time
Context: Up to 1M tokens; Model-dependent
File types: 4 ingest formats; PDF · DOCX · TXT · CSV
Inputs: Text · Image · PDF; Multimodal pipeline

Backend surfaces

Five extensions. One chat surface.

Multi-model-routing at send-time, long-term-memory across sessions, vision-input from images or PDFs. Every surface is wired to the AresGen backend with first-party endpoints.

Multi-model chat

AIChatController (core)

POST /dashboard/user/openai/chat/chat-send

Full models.tsx catalog routed at send-time

File chat

AIChatProFileChat (pipeline hook)

POST /dashboard/user/openai/chat/chat-send

PDF/DOCX/TXT/CSV embeddings (no own controller — hooks chatOutput)

Web chat

AIWebChat

POST /dashboard/user/openai/{locale}/webchat/chat-send

Live web-search augmentation

Memory & Folders

AIChatProMemory + AIChatProFolders

POST /ai-chat-pro-memory/save-instructions

User-level long-term context + folder taxonomy

Vision & Highlight

AiChatProSmartImage + AiChatProEntityHighlight

POST /dashboard/user/smart-image/upload

OpenAI Vision · Anthropic Vision · entity-detail / entity-image augmentation

Family coverage

Five frontier families. Multi-model-routing at send-time.

Pick a model from the rail or let the router decide. The catalog spans OpenAI, Anthropic, Google, Meta and Mistral. Every chat surface inherits the full set.

OpenAIAnthropicGoogleMetaMistral

Browse the full LLM catalog at /models.

Capabilities

Honest gaps. No marketing fog.

Long-term-memory crosses every surface; vision-input is scoped to two; web-search lives only in Web chat. The matrix below is generated from the same facts module the audit gate enforces.

Capability	Multi-model chat	File chat	Web chat	Memory & Folders	Vision & Highlight
Realtime streaming	✓	✓	✓	✓	✓
Vision input	✓	—	—	—	✓
File upload	—	✓	—	—	✓
Web search	—	—	✓	—	—
Long-term memory	✓	✓	✓	✓	✓
REST API	✓	✓	✓	✓	✓

Model rail

Twelve models. Five families. One subscription.

Filter the rail by family or scan the whole set. Every chip here is a real entry in the AresGen model catalog. Fabricated names fail CI.

GPT-5
OpenAI
GPT-5 Codex
OpenAI
GPT-4o
OpenAI
GPT-4.1
OpenAI
Claude Opus 4.7
Anthropic
Claude Sonnet 4.6
Anthropic
Claude Haiku 4.5
Anthropic
Gemini Pro 2.5
Google
Gemini Flash 2.5
Google
Llama 3.3 70B
Meta
Llama 3.1 405B
Meta
Mistral Large 2
Mistral

Pricing

One subscription. Every chat surface.

Lite

Try multi-model chat

Pro

File chat + memory + folders

Business

Web chat + vision + REST API

Custom

Volume + dedicated provider keys

Pro+ unlocks file chat, memory and folders. Business adds Web chat, vision and REST API. See /pricing for the breakdown.

Pair this with

Voice Studio

Add voice mode to any conversation.

Explore Voice Studio

Chatbots Studio

Deploy your chat as an embeddable agent.

Explore Chatbots

Use cases

Five workflows our customers ship every day.

Compare a Claude refactor against a GPT-4o variant in one conversation without losing context.

FAQ

Eight answers before you ask.

60+ frontier models including GPT-4o, Claude Opus/Sonnet/Haiku, Gemini 2.5 Pro, Llama 3.3, Mistral Large. Pick from the rail or let the router decide.

Stop juggling chat tabs.

Every frontier model, files, web search, vision, memory, in one subscription.

Start Free See pricing