A code-aware chat surface inside AresGen. Pick from twelve routed text models, paste a snippet, and use one of five canned prompts (Explain, Fix, Refactor, Add tests, Convert) or write your own.

Which model should I use for code?

For deep reasoning and large refactors, Claude Opus 4.7 or DeepSeek R1. For fast iteration, Claude Haiku 4.5 or Gemini Flash 2.5. For code-generation focus, GPT-5 Codex. The model rail surfaces per-task hints next to each chip.

Can AI Code read my whole repository?

No. AI Code accepts a single snippet upload per message. For codebase-wide context, paste the relevant file(s) into the chat. The single snippet upload surface is designed for focused review and refactor questions, not whole-repository indexing.

Can I switch models mid-conversation?

Yes. The MultiModel rail lets you swap models per prompt while keeping the conversation context. Try a fast model first, then route a follow-up to a stronger model when reasoning gets harder.

How are credits charged?

Each routed model deducts credits at its catalog tier. See the pricing page for tier breakdowns; there are no per-provider surcharges or separate API keys.

How is this different from AI Chat?

AI Code is the same routed-model conversation surface tuned for code: pre-built canned actions, markdown code-block rendering, and snippet upload front-and-center. AI Chat is the general-purpose entry point.

AresGen Tools

AI that reads your code in the model you choose.

Paste a snippet, pick a routed model, get a focused answer. Twelve text models across six families, with no manual API juggling and no provider lock-in.

Start free See pricing

12 routed models
6 families
Per-prompt switch
Markdown rendering

Routed models

Anthropic, OpenAI, Google, DeepSeek, Meta, Mistral

Provider families

Switch per prompt without changing tools

Backend surfaces

MultiModel rail + AIChatPro chat + single-file upload

Canned actions

Explain · Fix · Refactor · Add tests · Convert

Twelve models routed across six families.

Pick the model that suits the task and switch any time without losing the conversation.

Claude Opus 4.7Anthropic
Hard reasoning, long refactors
Claude Sonnet 4.6Anthropic
Balanced reasoning, daily code review
Claude Haiku 4.5Anthropic
Fast, short edits and snippets
GPT-5OpenAI
General-purpose coding
GPT-5 CodexOpenAI
Code generation focus
GPT-4.1OpenAI
Stable code edits
Gemini Pro 2.5Google
Long-context code reads
Gemini Flash 2.5Google
Quick lookups and scaffolds
DeepSeek V3DeepSeek
Strong coding benchmarks
DeepSeek R1DeepSeek
Reasoning-first code tasks
Llama 3.3 70BMeta
Open-weight baseline
Mistral Large 2Mistral
Reasoning + tool use

Family roster: Anthropic · OpenAI · Google · DeepSeek · Meta · Mistral.

Five canned actions, one click each.

Prompt templates for the most common code tasks. Click a chip to preview the kind of output you can expect.

Prompt: Explain what this code does step by step.

Before

function debounce(fn, ms) {
  let t;
  return (...a) => {
    clearTimeout(t);
    t = setTimeout(() => fn(...a), ms);
  };
}

After

// debounce returns a wrapper that delays calling `fn`
// until `ms` milliseconds have passed without a new call.
// Useful for keystroke and resize handlers.

Showcase output: copy-only, no execution. Generated content varies per model and prompt.

Three backend surfaces, one frontend layer.

Each surface ships a focused capability. Together they cover the common code workflow inside AresGen.

MultiModel rail

Switch between twelve text models in the same conversation. Pick the model that suits the task in front of you.

AIChatPro chat

A focused chat surface with markdown code-block rendering, syntax-highlighted output, and copy-ready snippets.

Single-file snippet upload

Upload one file per message: paste a function, a script, or a config and ask focused questions about it.

Five canned actions

Pre-built prompt templates for the most common code tasks: explain, fix, refactor, add tests, or convert across languages.

Six provider families.

Anthropic
Claude Opus 4.7, Sonnet 4.6, Haiku 4.5
OpenAI
GPT-5, GPT-5 Codex, GPT-4.1
Google
Gemini Pro 2.5, Flash 2.5
DeepSeek
DeepSeek V3, R1
Meta
Llama 3.3 70B
Mistral
Mistral Large 2

Pay per model, not per provider.

Each routed model deducts credits at its catalog tier. No separate API keys, no overage surprises. See the pricing page for tier breakdowns.

View pricing

Where AI Code fits.

Explain legacy code
Paste an unfamiliar function and route to Claude Opus 4.7 for a step-by-step walk-through with naming conventions intact.
PR review prep
Drop a diff into the chat and have GPT-5 Codex flag readability concerns before you hit the reviewer queue.
Refactor migration
Route to DeepSeek R1 for a focused refactor pass that swaps var/for loops for modern syntax without changing behavior.
Cross-language port
Paste a Python helper and ask Gemini Pro 2.5 to port it to TypeScript or Go, keeping types explicit.

Frequently asked questions.

What is AI Code?
A code-aware chat surface inside AresGen. Pick from twelve routed text models, paste a snippet, and use one of five canned prompts (Explain, Fix, Refactor, Add tests, Convert) or write your own.
Which model should I use for code?
For deep reasoning and large refactors, Claude Opus 4.7 or DeepSeek R1. For fast iteration, Claude Haiku 4.5 or Gemini Flash 2.5. For code-generation focus, GPT-5 Codex. The model rail surfaces per-task hints next to each chip.
Can AI Code read my whole repository?
No. AI Code accepts a single snippet upload per message. For codebase-wide context, paste the relevant file(s) into the chat. The single snippet upload surface is designed for focused review and refactor questions, not whole-repository indexing.
Can I switch models mid-conversation?
Yes. The MultiModel rail lets you swap models per prompt while keeping the conversation context. Try a fast model first, then route a follow-up to a stronger model when reasoning gets harder.
How are credits charged?
Each routed model deducts credits at its catalog tier. See the pricing page for tier breakdowns; there are no per-provider surcharges or separate API keys.
How is this different from AI Chat?
AI Code is the same routed-model conversation surface tuned for code: pre-built canned actions, markdown code-block rendering, and snippet upload front-and-center. AI Chat is the general-purpose entry point.