Coding Agents · Comparison · No score yet
Claude Code vs Codex
A fact sheet for builders. ShipRanked has not published test scores for Claude Code or Codex, and this page does not name a winner.
In one paragraph
ShipRanked has not published a test score for Claude Code or Codex, so this comparison does not name a winner. Claude Code runs as Terminal and desktop. Public price: Included with Claude Pro at $20/mo ($17/mo annual). Codex runs as Terminal, IDE, desktop, and web. Public price: Included with ChatGPT plans. API-key use is usage-based. Both are queued for “Add a feature and tests” and “Fix a seeded bug”. That queue is not a ranking. A published score is the median of three runs, from 0 to 100, and the top 3 stays empty until those runs are fresh.
Side by side
Facts checked October 15, 2026. Best-fit is not a verdict. Both tools are still in the test queue.
Price
Claude Code. Included with Claude Pro at $20/mo ($17/mo annual)
Codex. Included with ChatGPT plans. API-key use is usage-based.
Where it runs
Claude Code. Terminal and desktop
Codex. Terminal, IDE, desktop, and web
Model choice
Claude Code. Claude models. You switch with /model among the Fable, Opus, Sonnet, and Haiku aliases. On the Anthropic API those resolve to Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5. Console, Bedrock, Foundry, and Google Cloud still serve Claude models. This is not a bring-any-provider tool.
Codex. OpenAI models. The pricing page lists GPT-6.1 Sol, GPT-6 Astra, GPT-6 Sol, GPT-6 Luna, GPT-5.6 Sol, GPT-5.6 Luna, and GPT-5.5. The default depends on the CLI or IDE version. Sign in with ChatGPT, or use an API key. Not a bring-any-provider tool.
Open source
Claude Code. Closed source
Codex. The CLI is Apache-2.0. Hosted Codex is included with ChatGPT plans.
Best-fit job
Claude Code. Not measured. Queued for Add a feature and tests; Fix a seeded bug.
Codex. Not measured. Queued for Add a feature and tests; Fix a seeded bug.
Where to read the queue
Add a feature plus tests to a seeded repo
The agent gets the TaskBoard repo at a pinned commit and a written feature spec. It may edit application code and add tests. A hidden suite the agent does not see decides the pass.
Fix a seeded bug in a repository
A known bug is planted in TaskBoard. The agent has to find and fix it. The rest of the suite has to stay green, and the fix should not be a rewrite.