Model comparison
GPT-5.3-Codex vs GPT-5.2-Codex: same price, what differs
GPT-5.3-Codex and GPT-5.2-Codex cost the same to the cent. What differs is what OpenAI claims for each and where you can still run them in 2026.
· Prices as of September 28, 2026
GPT-5.3-Codex
OpenAI · Released February 5, 2026 · Previous generation
The February 2026 Codex-tuned coding model, which combined GPT-5.2-Codex's coding with stronger reasoning.
GPT-5.3-Codex facts and comparisonsGPT-5.2-Codex
OpenAI · Released December 18, 2025 · Previous generation
The December 2025 Codex-tuned version of GPT-5.2 for long-horizon agentic coding, refactors, and migrations.
GPT-5.2-Codex facts and comparisons
The short answer
GPT-5.3-Codex and GPT-5.2-Codex have identical rates, so every workload costs the same: $1.93 for the example agentic coding session on either. The choice comes down to OpenAI's claims, where GPT-5.3-Codex adds stronger reasoning and runs 25% faster for Codex users, and to availability. Both have left Codex for ChatGPT sign-in, and GPT-5.2-Codex has also left GitHub Copilot.
Choose GPT-5.3-Codex if
- You want the newer of the two, which OpenAI says combines GPT-5.2-Codex's coding with stronger reasoning and professional knowledge.
- Speed in Codex matters to you, and OpenAI says GPT-5.3-Codex runs 25% faster for Codex users.
- You use OpenCode or GitHub Copilot, which list GPT-5.3-Codex but not GPT-5.2-Codex.
Choose GPT-5.2-Codex if
- You have a pipeline validated on GPT-5.2-Codex in the API, where it remains available.
- Your work runs in Windows environments, where OpenAI claimed better performance for GPT-5.2-Codex.
- You lean on large refactors and migrations, which OpenAI named as GPT-5.2-Codex's focus.
Side by side
Specs and prices
| Fact | GPT-5.3-Codex | GPT-5.2-Codex |
|---|---|---|
| Maker | OpenAI | OpenAI |
| API model id | gpt-5.3-codex | gpt-5.2-codex |
| Released | February 5, 2026 | December 18, 2025 |
| Status | Previous generation | Previous generation |
| Context window | 400K tokens | 400K tokens |
| Max output | 128K tokens | 128K tokens |
| Open weights | No | No |
| Input, per 1M tokens | $1.75 | $1.75 |
| Cache hit, per 1M | $0.175 | $0.175 |
| Cache write, per 1M | $1.75 (same as input) | $1.75 (same as input) |
| Output, per 1M tokens | $14 | $14 |
| Runs in | Codex, Cursor, OpenCode, OpenRouter, and GitHub Copilot | Codex, Cursor, and OpenRouter |
Standard API rates in US dollars, as published by each maker on September 28, 2026. Batch and priority tiers, taxes, and subscription plans are not included.
Cost
What the same work costs
The same token counts, priced at each model’s published rates. Cached tokens are billed at each maker’s cache prices, so the session shows what caching is worth on each model.
Real sessions differ: the two models count the same code as different numbers of tokens, and reasoning settings change how much each one writes. Your own history is the real test.
| Workload | GPT-5.3-Codex | GPT-5.2-Codex |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $1.93 | $1.93 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.40 | $0.40 |
| Output-heavy generation, 30K input, 80K output | $1.17 | $1.17 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $211.75 | $211.75 |
| Where the session’s cost goes | ||
| Cache writes | $0.70 | $0.70 |
| Cache reads | $0.35 | $0.35 |
| Uncached input | $0.18 | $0.18 |
| Output | $0.70 | $0.70 |
- caching saves on the session with GPT-5.3-Codex (62%)
- $3.15
- caching saves on the session with GPT-5.2-Codex (62%)
- $3.15
Do GPT-5.3-Codex and GPT-5.2-Codex cost the same?
To the cent. GPT-5.3-Codex and GPT-5.2-Codex both charge $1.75 per million input tokens, $0.175 per million cached tokens, and $14 per million output tokens. Both predate GPT-5.6, so written cache tokens cost ordinary input and hits cost 0.1x.
Every example workload matches. The session costs $1.93, the uncached review $0.40, and the output-heavy generation $1.17 on either model. A month of 110 sessions comes to $211.75 on both, and caching saves $3.15, or 62%, on each.
Output carries unusual weight here. At $14 per million, output is 36% of the session, as large a share as the cache writes. That makes reasoning length the lever that could separate the two in practice, and fixed workloads can't show it. OpenAI says GPT-5.3-Codex reasons more strongly, but whether it writes more or fewer tokens on your tasks is something only your own history shows.
What OpenAI changed in GPT-5.3-Codex
OpenAI released GPT-5.2-Codex in December 2025 as a Codex-tuned version of GPT-5.2 for long-horizon agentic coding, and called it its most intelligent coding model optimized for long-horizon, agentic coding tasks. It highlighted context compaction for long work, large refactors and migrations, and better performance in Windows environments.
GPT-5.3-Codex followed in February 2026. OpenAI said it combined GPT-5.2-Codex's coding with stronger reasoning and professional knowledge, ran 25% faster for Codex users, and collaborated better while the agent was working. At launch OpenAI called it the most capable agentic coding model to date.
Both share a 400K context window, accept up to 272K input tokens of it, and write up to 128K tokens of output. Neither has a long-context price tier, because neither accepts input past 272K.
Where you can still run each Codex model
Neither can be picked in Codex with ChatGPT sign-in anymore. GPT-5.2-Codex was removed on April 14, 2026, and GPT-5.3-Codex stopped being selectable on May 26, 2026. Both remain in the OpenAI API, and API-key use of GPT-5.3-Codex in Codex is unaffected.
Outside Codex they differ more. GPT-5.2-Codex was retired from GitHub Copilot on June 5, 2026, and is not listed in OpenCode, while GPT-5.3-Codex is still listed in both. Each remains available in Cursor and on OpenRouter.
Neither is where OpenAI's current recommendations point. The Codex docs recommend GPT-6 Sol for complex coding, and it follows newer caching rules with a 1.25x cache-write charge. If you still run one of these two with an API key, EveryToken prices each local Codex request at API rates, which gives you a baseline before any move.
Prompt caching
How OpenAI bills cached tokens
OpenAI
Prompt caching is on by default. From GPT-5.6 on you can also mark up to four explicit cache breakpoints, while GPT-5.5 and earlier cache automatically only.
From GPT-5.6 on, a cache write costs 1.25x the uncached input price and a cache hit costs 0.1x. GPT-5.5 and earlier add no charge for writing the cache: written tokens are billed as ordinary input, and a hit costs 0.1x on the models compared here.
On GPT-5.6 and later, a cached prefix stays reusable for at least 30 minutes after its last use, and caching starts at 1,024 input tokens.
Source: OpenAI: Prompt caching
Your own numbers
See what GPT-5.3-Codex and GPT-5.2-Codex really cost you.
everyaitoken reads your Codex, Cursor, OpenCode, and OpenRouter history on your Mac and prices every request at API rates, with what caching saved or cost. $9 once.
FAQ
Questions
Is GPT-5.3-Codex more expensive than GPT-5.2-Codex?
No. The rates are identical, $1.75 for input, $0.175 for cache hits, and $14 for output per million tokens, so every workload on this page costs the same on both.
Is GPT-5.2-Codex still available?
Yes, in the OpenAI API, in Cursor, and on OpenRouter. It was removed from Codex for ChatGPT sign-in on April 14, 2026, and retired from GitHub Copilot on June 5, 2026.
What is the context limit of GPT-5.3-Codex and GPT-5.2-Codex?
Both have a 400K context window and accept up to 272K input tokens of it. Both write up to 128K tokens of output.
What does GPT-5.3-Codex add over GPT-5.2-Codex?
OpenAI says it combines GPT-5.2-Codex's coding with stronger reasoning and professional knowledge and runs 25% faster for Codex users. It also says GPT-5.3-Codex collaborates better while the agent is working.