Model comparison
GPT-5.3-Codex vs Gemini 3.1 Pro Preview: context vs cost
GPT-5.3-Codex and Gemini 3.1 Pro Preview cost within 4% of each other on a cached coding session. Context size, output limits, and access set them apart.
· 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 comparisonsGemini 3.1 Pro Preview
Google · Released February 19, 2026 · Preview
Google's only current Pro model, still in preview, positioned for deep reasoning and agentic coding.
Gemini 3.1 Pro Preview facts and comparisons
The short answer
Price barely separates them: the example agentic coding session costs $1.93 on GPT-5.3-Codex and $2.00 on Gemini 3.1 Pro Preview, while output-heavy work is cheaper on Gemini at $1.02 against $1.17. Limits matter more, since GPT-5.3-Codex accepts at most 272K input tokens and writes up to 128K, while Gemini 3.1 Pro Preview takes a 1.05M window but writes at most 65.5K. Pick GPT-5.3-Codex for existing API-key Codex workflows and long outputs, and Gemini 3.1 Pro Preview for very large prompts in Gemini CLI.
Choose GPT-5.3-Codex if
- You need long outputs, up to 128K tokens against 65.5K.
- You already run it in Codex with an API key, which its removal from ChatGPT sign-in on May 26, 2026 did not affect.
- You work in GitHub Copilot, which lists GPT-5.3-Codex and retired Gemini 3.1 Pro Preview on September 1, 2026.
- Your sessions are input-heavy, where $1.75 input and $0.175 cache hits undercut Gemini's $2 and $0.20.
Choose Gemini 3.1 Pro Preview if
- You need more than 272K input tokens in one request, up to a 1.05M window, and accept higher rates above 200K.
- Your work is output-heavy, where $12 per million output tokens undercuts GPT-5.3-Codex's $14.
- Gemini CLI is your tool, and it already uses this model as the Pro half of its default auto model.
Side by side
Specs and prices
| Fact | GPT-5.3-Codex | Gemini 3.1 Pro Preview |
|---|---|---|
| Maker | OpenAI | |
| API model id | gpt-5.3-codex | gemini-3.1-pro-preview |
| Released | February 5, 2026 | February 19, 2026 |
| Status | Previous generation | Preview |
| Context window | 400K tokens | 1.05M tokens |
| Max output | 128K tokens | 65.5K tokens |
| Open weights | No | No |
| Input, per 1M tokens | $1.75 | $2 |
| Cache hit, per 1M | $0.175 | $0.20 |
| Cache write, per 1M | $1.75 (same as input) | $2 (same as input) |
| Output, per 1M tokens | $14 | $12 |
| Runs in | Codex, Cursor, OpenCode, OpenRouter, and GitHub Copilot | Cursor, Gemini CLI, OpenCode, 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. Gemini 3.1 Pro Preview: Prompts over 200K input tokens cost $4 input, $0.40 cached, and $18 output per million tokens.
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 | Gemini 3.1 Pro Preview |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $1.93 | $2.00 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.40 | $0.42 |
| Output-heavy generation, 30K input, 80K output | $1.17 | $1.02 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $211.75 | $220.00 |
| Where the session’s cost goes | ||
| Cache writes | $0.70 | $0.80 |
| Cache reads | $0.35 | $0.40 |
| Uncached input | $0.18 | $0.20 |
| Output | $0.70 | $0.60 |
- caching saves on the session with GPT-5.3-Codex (62%)
- $3.15
- caching saves on the session with Gemini 3.1 Pro Preview (64%)
- $3.60
Two February 2026 models priced within cents
OpenAI released GPT-5.3-Codex on February 5, 2026, and Google released Gemini 3.1 Pro Preview two weeks later, on February 19. Their prices sit close together: $1.75 against $2 for input, $0.175 against $0.20 for a cache hit, and $14 against $12 for output per million tokens.
Neither charges a premium to write the cache. OpenAI bills written tokens as ordinary input on GPT-5.5 and earlier models, and Google publishes no separate write price. Both charge 10% of input for a hit, which leaves input and output prices to decide each workload.
On the example session GPT-5.3-Codex is $0.07 cheaper, $1.93 against $2.00, and on the uncached review $0.02 cheaper, $0.40 against $0.42. On output-heavy generation Gemini is $0.15 cheaper, $1.02 against $1.17. Output makes up 36% of GPT-5.3-Codex's session cost and 30% of Gemini's, so the more a task writes, the more Gemini's lower output price counts. A month of 110 sessions comes to $211.75 against $220.00.
400K against 1.05M: how much context each accepts
Here the models differ sharply. GPT-5.3-Codex's context window is 400K, and at most 272K of that can be input. Gemini 3.1 Pro Preview's window is 1.05M, 2.6x larger. An agent that needs more than 272K tokens of code in a single request can only send it to Gemini of these two.
Large prompts cost more on Gemini, though. Above 200K input tokens Google charges $4 input, $0.40 cached, and $18 output per million, so a prompt between 200K and 272K tokens costs more on Gemini than on GPT-5.3-Codex, whose listed pricing has no long-context tier. On output the ceiling runs the other way: GPT-5.3-Codex writes up to 128K tokens and Gemini at most 65.5K.
Access: legacy in Codex, preview at Google
GPT-5.3-Codex has not been selectable in Codex with ChatGPT sign-in since May 26, 2026, but API-key use is unaffected, and it is still listed in Cursor, OpenCode, OpenRouter, and GitHub Copilot. OpenAI's documentation calls it "The most capable agentic coding model to date," and its Codex changelog said it combined GPT-5.2-Codex's coding with stronger reasoning and professional knowledge, running 25% faster for Codex users. The Codex docs now recommend GPT-6 Sol for complex coding.
Gemini 3.1 Pro Preview is still a preview, with Gemini 3.5 Pro announced but not released. It is the Pro half of Gemini CLI's default auto model, and Gemini API key users of the CLI get its customtools endpoint at the same price. Its thinking cannot be turned off and defaults to high, which affects how many output tokens a task uses.
With the two this close on price, a few cents on a sample table matter less than your own mix of work. EveryToken shows what each model costs on your Codex and Gemini CLI sessions, priced at API rates with cache savings.
Prompt caching
How each maker 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
Implicit caching is on by default for Gemini 2.5 and newer. When a request repeats a prefix Google has cached, the discount is applied automatically, but a hit is not guaranteed.
Explicit caching creates a cache you reference by name, with a guaranteed discount. It adds a storage charge for as long as the cache lives, 1 hour by default: $4.50 per million tokens per hour on Pro models and $0.50 to $1 on Flash models.
On every model compared here, a cache hit costs 10% of the input price. Google publishes no separate cache-write price, so this blog prices written tokens as ordinary input.
Source: Google: Context caching
Your own numbers
See what GPT-5.3-Codex and Gemini 3.1 Pro Preview really cost you.
everyaitoken reads your Codex, Cursor, OpenCode, OpenRouter, and Gemini CLI 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 cheaper than Gemini 3.1 Pro Preview?
Slightly, on input-heavy work: the example agentic session costs $1.93 against $2.00, 4% less. Gemini is cheaper on output-heavy generation, $1.02 against $1.17, because its output price is $12 against $14. Both figures are API-equivalent estimates.
Can I still use GPT-5.3-Codex in Codex?
Only with an API key. It has not been selectable with ChatGPT sign-in since May 26, 2026, and API-key use is unaffected.
Which model takes a larger prompt?
Gemini 3.1 Pro Preview, with a 1.05M context window. GPT-5.3-Codex's window is 400K, of which it accepts up to 272K as input.
Do either of these models charge for cache writes?
No. Both bill written tokens at their ordinary input price, $1.75 on GPT-5.3-Codex and $2 on Gemini 3.1 Pro Preview. A cache hit costs 10% of input on each. Gemini's explicit caching adds storage at $4.50 per million tokens per hour on Pro models.
Sources
- OpenAI: API pricing
- OpenAI docs: GPT-5.3-Codex
- Codex docs: Changelog
- OpenAI: API changelog
- Codex docs: Models
- Cursor docs: Models and pricing
- OpenRouter: GPT-5.3-Codex
- Google: Gemini API pricing
- Google docs: Gemini 3.1 Pro Preview
- Google: Gemini models
- Google: Gemini 3.1 Pro
- Google DeepMind: Gemini
- Gemini CLI source: model configuration
- Cursor docs: Gemini 3.1 Pro
- OpenRouter: Gemini 3.1 Pro Preview
- OpenCode docs: Zen
- GitHub Docs: Supported AI models in Copilot
- OpenAI: Prompt caching
- Google: Context caching