Use GPT-5.3 Codex in Kilo Code
OpenAI's most powerful coding model unifies frontier reasoning and code generation. Plan in Architect mode, implement in Code mode, debug in Debug mode — one model across your entire workflow.
Two Ways to Use GPT-5.3 Codex in Kilo Code
Pay-as-you-go through Kilo Gateway or flat-rate with your ChatGPT subscription
Kilo Gateway
Pay-as-you-go with zero markup on API token usage. Access GPT-5.3 Codex across all Kilo features including Cloud agents.
Boost your credits with Kilo Pass — get up to +50% bonus API credits on your subscription.
Explore Kilo Gateway →ChatGPT Subscription (Login with Codex)
Use your existing ChatGPT Plus or Pro subscription for flat-rate access. OAuth login — no API keys needed. Full agentic workflows in your IDE.
Note: ChatGPT subscription path provides local IDE features only (no cloud agents).
Set Up Login with Codex →One Model, Every Mode
GPT-5.3 Codex handles the entire workflow — from planning to shipping
Architect
Plan
Analyze requirements, design system architecture, and create implementation plans with elite reasoning.
Code
Implement
Generate, refactor, and edit code with frontier-level accuracy and token efficiency.
Debug
Troubleshoot
Identify root causes, trace execution flows, and fix bugs with deep contextual understanding.
Orchestrator
Coordinate
Break down complex tasks, delegate to sub-agents, and manage multi-step workflows end to end.
Why GPT-5.3 Codex
Unified reasoning and coding in a single frontier model
Unified Reasoning & Coding
Merges frontier code generation from the Codex line with elite professional reasoning from the GPT-5.2 Thinking series. One model that plans and executes.
25% Faster & More Efficient
Approximately 25% faster and more token-efficient than prior Codex models. Faster responses, lower costs, and higher throughput on complex tasks.
Benchmark Leader
77.3% Terminal-Bench 2.0, 56.8% SWE-Bench Pro, 64.7% OSWorld-Verified, 70.9% GDPval — a massive jump from prior generations across the board.
Strong Compaction Support
Efficiently compresses and manages context across long coding sessions. Maintains coherence and accuracy even as conversations grow complex.
Long-Running Autonomy
Built for sustained autonomous workflows — can tackle multi-step tasks over extended sessions without losing track of goals or context.
Built for Agentic Workflows
Instrumental in creating itself — OpenAI engineers used it to debug training runs. Output is more reliable, with warmth and willingness to just do things.
GPT-5.3 Codex Benchmark Performance
A massive leap in agentic coding — 77.3% on Terminal-Bench 2.0
Read the full blog post →| Benchmark | GPT-5.3 Codex | GPT-5.2 Codex | Opus 4.6 | Gemini 3.1 Pro |
|---|---|---|---|---|
| Terminal-Bench 2.0 | 77.3% | 64.0% | 65.4% | 56.2% |
| SWE-Bench Pro | 56.8% | — | 56.8% | 54.2% |
| OSWorld-Verified | 64.7% | — | 72.7% | — |
| GDPval (Knowledge Work) | 70.9% | — | — | — |
Get Started with GPT-5.3 Codex
Up and running in minutes
Install Kilo Code
Add Kilo Code to VS Code, JetBrains IDE, or use the CLI.
Choose your path
Use Kilo Gateway (pay-as-you-go) or log in with your ChatGPT Plus/Pro subscription.
Select GPT-5.3 Codex
Pick GPT-5.3 Codex from the model picker and start building.
Trusted by developers at the world's most innovative companies
Agentic Engineering
Glide through your workflow with a mode for every step
Ask mode
A knowledgeable technical assistant focused on answering questions without changing your codebase
Use Kilo Everywhere
Kilo works where you work. Build alone or with your team.
Frequently Asked Questions
Compare Kilo Code with Other Tools
See how Kilo Code stacks up against other AI coding assistants
Kilo Code vs Cursor
The open-source agentic platform inside your existing IDE vs the standalone AI-first code editor. 500+ models, zero markup, no editor switch required.
Kilo Code vs GitHub Copilot
Multi-model, multi-mode agentic coding vs a single-vendor autocomplete from GitHub.
Kilo Code vs Windsurf
500+ models at exact provider rates. No credit system. Full BYOK on all plans. Open source.
Kilo Code vs Claude Code
Open-source, multi-model CLI + IDE agent with inline autocomplete vs Anthropic's Claude-only terminal-first coding agent.
Kilo Code vs Roo Code
Roo Code archives May 15, 2026 — compare it with Kilo Code and migrate in minutes.
Kilo Code vs Cline
Kilo Code bundles Cline-style autonomy plus Orchestrator, Architect, Debug, and Code modes.
Kilo Code vs Tabnine
Full agentic coding across 500+ models vs Tabnine's enterprise-only autocomplete.
Kilo Code vs Augment Code
Open-source, BYOK-everywhere agent platform vs Augment Code's closed proprietary stack.
Kilo Code vs Lovable
Kilo Code plugs into your real IDE and codebase — Lovable is a hosted AI app builder optimized for greenfield prototyping.
Kilo Code vs Replit
Use Kilo inside your existing editor and infrastructure vs Replit's hosted browser IDE.
Kilo Code vs Warp
Kilo Code works inside VS Code and JetBrains as an AI coding agent. Warp is a standalone terminal-first Agentic Development Environment. Same agent power, different home.
Kilo Code vs Amp Code
Amp Code is shutting down its VS Code extension — compare it with Kilo Code
Kilo Code for JetBrains IDEs
AI coding assistant for IntelliJ, PhpStorm, WebStorm, and Rider
Ready to Code with GPT-5.3 Codex?
Join 3M+ developers using Kilo Code. One model that reasons, codes, and debugs — across every mode.