Short answer
For a production API decision, GPT-5.2 and Gemini 3 Pro are no longer equivalent live options. Google says the gemini-3-pro-preview model on the Gemini API was shut down on March 9, 2026. OpenAI still documents the gpt-5.2 API model and its dated snapshot, while describing GPT-5.2 as a previous flagship. That makes Gemini 3 Pro a historical API comparison, not a current endpoint to select. This statement concerns the documented Gemini API model; it does not claim that every Google product or channel has the same status. (ai.google.dev)
Historically, Gemini 3 Pro Preview had a larger input limit and accepted more input modalities. GPT-5.2 has documented text and image input, text output, reasoning settings, API features, and token rates. Those specifications can help explain older comparisons, but they do not make the shut-down Gemini API endpoint usable today. (developers.openai.com)
What the API comparison actually covers
| OpenAI GPT-5.2 API | Google Gemini 3 Pro Preview API | |
|---|---|---|
| Model ID | gpt-5.2; dated snapshot gpt-5.2-2025-12-11 |
gemini-3-pro-preview |
| API status | Documented as a previous flagship; OpenAI lists the model ID and snapshot | Shut down March 9, 2026 |
| Input and output | Text and image input; text output | Historically: text, image, video, audio, and PDF input; text output |
| Context and output limits | 400,000-token context; maximum 128,000 output tokens | Historical limits: up to 1,048,576 input tokens and 65,536 output tokens |
| Selected API features | Streaming, function calling, structured outputs, and reasoning-effort settings are documented | Historical model page lists function calling, structured outputs, thinking, code execution, search grounding, and URL context |
| Token pricing | OpenAI lists $1.75 per million input tokens, $0.175 per million cached input tokens, and $14 per million output tokens | No current price for a usable Gemini 3 Pro API endpoint. At launch, Google listed $2 per million input and $12 per million output tokens for prompts of 200,000 tokens or less. |
GPT-5.2 specifications and rates above are from OpenAI’s API model documentation. Gemini 3 Pro limits and features are historical details from Google’s model page; its launch prices are historical and apply to the stated prompt-size band. These are API details, not a comparison of ChatGPT and consumer Gemini subscriptions. (developers.openai.com)
Which one should you choose?
- For an existing GPT-5.2 API integration: OpenAI documents text and image input, text output, function calling, structured outputs, streaming, and reasoning-effort settings from
nonethroughxhigh. Check the current model documentation and your account’s rate limits before relying on a particular integration or workload. (developers.openai.com) - For a new Google API integration: Do not select
gemini-3-pro-preview; Google documents that this Gemini API model was shut down. Consult Google’s current model catalog for available choices instead. (ai.google.dev) - For a new OpenAI API integration: GPT-5.2 is documented as a previous flagship, so compare it with the models currently listed in OpenAI’s catalog rather than assuming it is the best fit for a new deployment. (
developers.openai.com)
Older comparison articles sometimes describe Gemini 3 Pro as a candidate for very long-context or multimodal work. Its historical input window and modalities help explain that positioning, but a shut-down API model cannot be chosen for a new production deployment through that endpoint.
What published comparisons do—and don’t—prove
A Kilo AI article published December 12, 2025, describes three custom coding tasks: a tightly specified Python rate limiter, a TypeScript API-handler refactor, and an extension to a notification system. The article reports that Gemini 3.0 did well on the rate-limiter prompt, while GPT-5.2 was more complete on the two more complex tasks. It does not establish that its tested Gemini model was Gemini 3 Pro, and the results are limited to those prompts—not a general model ranking or official benchmark. The article also distinguishes GPT-5.2 from GPT-5.2 Pro, so its Pro results should not be attributed to GPT-5.2. (Kilo AI’s test report)
A separate comparison discusses GPT-5.2 versus Gemini 3, not a confirmed Gemini 3 Pro API variant, and provides no numerical GPQA score in the captured article text. Its qualitative claims should not be treated as verified specifications or direct evidence about Gemini 3 Pro. (Scalevise comparison)
GPT-5.2 API cost example
OpenAI lists GPT-5.2 at $1.75 per million uncached input tokens and $14 per million output tokens. For aggregate usage across several valid requests totaling 1 million uncached input tokens and 200,000 output tokens, the model-token charges would be:
(1 × $1.75) + (0.2 × $14) = $4.55
This is a calculation from the published per-token rates, not a measured workload or a complete bill. The totals are spread across requests rather than representing one request; each request must stay within the model’s applicable limits. The calculation excludes cached input, separately billed tools, and other charges. Gemini’s launch rates are not a like-for-like current price comparison because its Gemini API endpoint is shut down. (developers.openai.com)
Kilo AI’s reported spending for coding-agent runs is also not API list pricing: those figures refer to its custom tasks in Kilo Code. Do not compare them directly with provider token rates. (Kilo AI’s test report)
How to compare current production candidates
Compare model IDs that are available through the APIs you intend to use. Run representative requests from your own workload against the same success criteria—for example, required-field accuracy for extraction, test pass rate and requirement coverage for coding, or answer quality against a reference set for analysis. Record latency and token usage, review failures, and calculate cost using each provider’s current billing units. This is an evaluation method, not a test performed for this article.
Bottom line: GPT-5.2 versus Gemini 3 Pro is now mainly a historical API comparison. Google’s Gemini 3 Pro Preview API endpoint was shut down; OpenAI still documents GPT-5.2 as a previous flagship. For a production selection today, compare supported current model IDs and official prices rather than treating Gemini 3 Pro’s old preview specifications or launch rates as a live offer.
Sources
Gemini 3 Pro preview | Gemini API
Gemini 3 for developers: New reasoning, agentic capabilities
We Tested GPT-5.2/Pro vs. Opus 4.5 vs. Gemini 3 on 3 Real-World Coding Tasks