GPT-6.1 Sol vs GPT-5.3 Codex
Green marks the cheaper side per row — informational only, never a verdict for your workload
| Status | priceable | priceable |
|---|---|---|
| Context window | 1,050,000 tokens | 400,000 tokens |
| Knowledge cutoff | Apr 30, 2026 | Aug 31, 2025 |
| Standard input | $2.00/MTok | $1.75/MTok |
| Standard cached input | $0.10/MTok | $0.175/MTok |
| Standard output | $10.00/MTok | $14.00/MTok |
| Cache writes | Write rate $2.50/MTok | No surcharge |
| Reasoning effort levels | low, medium (default), high, xhigh, max | low, medium, high, xhigh |
| Faster tiers offered | Fast | Fast |
| Long-context repricing rule | Published | Not published |
| Tier 1 rate limit | 500 RPM · 500,000 TPM | 500 RPM · 500,000 TPM |
| ChatGPT credits per 1M (in / out) | 50 / 250 | No credit rate |
| Codex with ChatGPT sign-in | No retirement announced | Deprecated |
The Codex CLI's default against the model named for Codex. This pairing used to be the site's flagship question, and it's changed shape: GPT-5.3 Codex is now deprecated in Codex for anyone signed in with ChatGPT, which makes it a choice only API-key users can still make.
Where the default pulls ahead
Context window, by a wide margin — GPT-5.3 Codex has a fraction of GPT-6.1 Sol's window. Knowledge cutoff, by most of a year. Effort range, too: the default goes from low to max with medium as its default, while GPT-5.3 Codex stops at xhigh and marks no default. And service tiers — the default publishes Batch and Flex prices, while GPT-5.3 Codex publishes only Standard and Fast.
Where the Codex model is cheaper
Uncached input only. GPT-5.3 Codex charges less per uncached input token, but more per cached input token and per output token, and it publishes no cache-write rate. Agentic sessions are dominated by cached context and output, so the per-token input advantage rarely survives contact with a real session; price your own token mix with the token cost estimator.
What's unknown on one side
The default publishes the long-context repricing rule. GPT-5.3 Codex's page doesn't state it either way, so this site records it as not published.
What doesn't differ
Requests and tokens per minute match tier for tier. GPT-5.3 Codex additionally publishes batch-queue limits, though with no Batch price to use them with.
How to decide
For ChatGPT sign-in, use the default and don't build new workflows on GPT-5.3 Codex. With an API key, choose it only for a narrow workload where small context and cheap uncached input genuinely dominate — and read why the CLI defaults to the general model first.
Verified 2026-10-01 against CodexHow facts module (src/data/facts/) — see /about/#accuracy.