OpenAI: GPT-5 Image Mini vs Tencent: Hy Image 3.5 Preview
A side-by-side comparison of OpenAI: GPT-5 Image Mini ($8.75 / 1M tokens) and Tencent: Hy Image 3.5 Preview ($5.60 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.
Specifications
| OpenAI: GPT-5 Image Mini | Tencent: Hy Image 3.5 Preview | |
|---|---|---|
| Provider | OpenAI | Tencent |
| Input | $8.75 / 1M tokens | - |
| Output | $7.00 / 1M tokens | - |
| Cache read | $0.88 / 1M tokens | - |
| Image output | $28.00 / 1M tokens | $5.60 / 1M tokens |
| Web search | $35.00 / 1K searches | - |
| Image input | - | $5.60 / 1M tokens |
| Context window | 400,000 | 100,000 |
| Max output | 128,000 | 90,000 |
| Input | file, image, text | text, image |
| Output | image, text | image |
| Tool calling | ||
| Structured output | ||
| Reasoning control | ||
| Added | Oct 16, 2025 | Oct 5, 2026 |
Key differences
- OpenAI: GPT-5 Image Mini reads 4.0× more context (400,000 tokens).
- Only OpenAI: GPT-5 Image Mini supports structured JSON output.
- Only OpenAI: GPT-5 Image Mini supports reasoning effort control.
In depth
OpenAI: GPT-5 Image Mini is an image model from OpenAI, released on 16 October 2025. Tencent: Hy Image 3.5 Preview is an image model from Tencent, released on 5 October 2026. Both sit in the image category, so both are meant for work where pictures are either the input, the output, or both. The main difference at the top level is who runs them and how much surface area each one exposes to the caller.
The capability gap is wider than the naming suggests. GPT-5 Image Mini takes file, image and text as input and can return image and text as output, while Hy Image 3.5 Preview takes text and image as input and returns image only. GPT-5 Image Mini also carries a much larger context window at 400000 tokens against 100000 for the Tencent model, so it can hold far more material in a single request. On top of that, GPT-5 Image Mini is the only one of the two with structured output support and the only one with reasoning support, and it exposes a long list of parameters including temperature, top_p, max_tokens, seed, logprobs, response_format and verbosity. Hy Image 3.5 Preview lists no supported parameters at all, which means you get a much narrower control surface. The qualitative comparison marks GPT-5 Image Mini as the bigger context model, the only structured model and the only reasoning model. Neither side is flagged as cheaper on input or output, so treat them as about the same price rather than assuming one undercuts the other.
For practical selection, reach for GPT-5 Image Mini when the task needs more than a single image back. If you want the model to reason about what it is producing, return structured data alongside or instead of pixels, or accept a mix of text and image in the response, it is the only option here that can do that. Its 400000 token window also suits jobs where you are feeding in long documents, many reference images or a large file alongside the prompt. Hy Image 3.5 Preview is the simpler tool: text and image in, image out. If your pipeline only needs an image generated from a prompt plus optional reference image, and you do not need parameters, reasoning traces or structured payloads, it covers that case without extra machinery. Because pricing is not differentiated in the data, the decision should rest on modality and control needs rather than cost.
Calling either model through api14 works the same way because the gateway is OpenAI-compatible. You point your existing client at the api14 base URL, send a request in the usual chat completions shape, and change only the model id string. Use "openai/gpt-5-image-mini" when you want the OpenAI model and "tencent/hy-image-v3.5-preview" when you want the Tencent one. Everything else in the request body stays where it is, though you should remember that parameters such as temperature, seed or response_format will only take effect on the model that supports them, which in this pair is GPT-5 Image Mini.
There is one case where the Tencent model remains the better pick even though it is the more limited of the two. If your application is purely image generation from text and image input, and you have no use for reasoning, structured outputs or a large context window, then Hy Image 3.5 Preview matches the job exactly. Adding GPT-5 Image Mini there would give you capabilities you never call, so the narrower model is the cleaner fit for that specific workflow.
Switch with one string
Both models use the same endpoint and request shape. Change the model value from "openai/gpt-5-image-mini" to "tencent/hy-image-v3.5-preview" and prompts, streaming and tools keep working.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api14.com/api/v1",
apiKey: process.env.API14_API_KEY,
});
const result = await client.images.generate({
"model": "tencent/hy-image-v3.5-preview",
"prompt": "A minimalist product photo of a ceramic mug on a wooden table",
"n": 1,
"aspect_ratio": "1:1"
});
import { writeFile } from "node:fs/promises";
await writeFile("image.png", Buffer.from(result.data[0].b64_json, "base64"));
More comparisons
- OpenAI: GPT Image 2.6 Flare vs OpenAI: GPT Image 2.6 Sunburst
- OpenAI: GPT Image 2.6 Flare vs Auto Router
- OpenAI: GPT Image 2.6 Flare vs Auto Code
- OpenAI: GPT Image 2.6 Flare vs Sourceful: Riverflow V2.6 Pro
- OpenAI: GPT Image 2.6 Sunburst vs Auto Router
- OpenAI: GPT Image 2.6 Sunburst vs Auto Code
- OpenAI: GPT Image 2.6 Sunburst vs Sourceful: Riverflow V2.6 Pro
- Auto Router vs Auto Code