We use cookies. Cookies and browser storage keep you signed in and remember your preferences. With your permission we also measure how the site is used. You can any time. See our cookie policy.

New here? See how it works A 1-minute tour of the playground, models and your API key.

Gemini 3.1 Flash Lite Preview vs Qwen2.5 VL 72B Instruct: API price, context & features · api14

Google: Gemini 3.1 Flash Lite Preview vs Qwen: Qwen2.5 VL 72B Instruct

A side-by-side comparison of Google: Gemini 3.1 Flash Lite Preview ($0.88 / 1M tokens) and Qwen: Qwen2.5 VL 72B Instruct ($2.80 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.

Specifications

	Google: Gemini 3.1 Flash Lite Preview	Qwen: Qwen2.5 VL 72B Instruct
Provider	Google	Qwen
Input	$0.88 / 1M tokens	$2.80 / 1M tokens
Output	$5.25 / 1M tokens	$3.50 / 1M tokens
Cache read	$0.09 / 1M tokens	$1.40 / 1M tokens
Cache write	$0.29 / 1M tokens	-
Reasoning	$5.25 / 1M tokens	-
Audio input	$1.75 / 1M tokens	-
Image	$0.000001 / image	-
Web search	$49.00 / 1K searches	-
Context window	1,048,576	128,000
Max output	65,536	115,200
Input	text, image, video, file, audio	text, image
Output	text	text
Tool calling
Structured output
Reasoning control
Added	Mar 3, 2026	Feb 1, 2025

Key differences

Google: Gemini 3.1 Flash Lite Preview is 69% cheaper on input tokens.
Qwen: Qwen2.5 VL 72B Instruct is 33% cheaper on output tokens.
Google: Gemini 3.1 Flash Lite Preview reads 8.2× more context (1,048,576 tokens).
Only Google: Gemini 3.1 Flash Lite Preview supports tool calling.
Only Google: Gemini 3.1 Flash Lite Preview supports reasoning effort control.
A request with 1,000 input and 500 output tokens costs $0.00350 with Google: Gemini 3.1 Flash Lite Preview and $0.00455 with Qwen: Qwen2.5 VL 72B Instruct.

In depth

Google: Gemini 3.1 Flash Lite Preview is a Google model, listed under the text category, and it first appeared from the provider on 2026-03-03. Qwen: Qwen2.5 VL 72B Instruct comes from Qwen, also sits in the text category, and carries a provider release date of 2025-02-01. Both are served through api14 as text output models, so the choice between them comes down to how each one handles input and how much room it gives you.

The biggest split is context and input modality. Gemini 3.1 Flash Lite Preview accepts text, image, video, file and audio input and offers a context window of 1,048,576 tokens, which is the larger of the two. Qwen2.5 VL 72B Instruct accepts text and image only, with a context window of 128,000 tokens. Gemini 3.1 Flash Lite Preview is also the only one of the pair that exposes tool calling and reasoning controls, since its parameter list includes tools, tool_choice, reasoning, reasoning_effort and include_reasoning. On structured outputs the two are not meaningfully different, and both let you set response_format, temperature, top_p, max_tokens and seed. Qwen2.5 VL 72B Instruct instead gives you a wider set of sampling and inspection knobs, including frequency_penalty, presence_penalty, repetition_penalty, logit_bias, logprobs, top_logprobs, top_k and stop. On price, Gemini 3.1 Flash Lite Preview is the cheaper option for input, while Qwen2.5 VL 72B Instruct is the cheaper option for output.

For practical work, reach for Gemini 3.1 Flash Lite Preview when the job involves long documents, large codebases or mixed media, since it can take video, audio and files alongside text and has far more context to work with. It is also the pick when you need the model to call tools or when you want to turn reasoning on and tune how much of it you get. Choose Qwen2.5 VL 72B Instruct when your input is text plus images and 128,000 tokens is enough, and when you care about output cost or want finer control over sampling through penalties, top_k, logit_bias and log probability outputs. If your workload is mostly image understanding with short prompts and you are watching generation spend, Qwen2.5 VL 72B Instruct remains the better fit even though Gemini 3.1 Flash Lite Preview is cheaper on input.

Calling either model through api14 works the same way, because the API is OpenAI compatible. You point your client at the same endpoint, send the same chat completions style request body, and change only the model id. Use google/gemini-3.1-flash-lite-preview for the Google model and qwen/qwen2.5-vl-72b-instruct for the Qwen model. Everything else, your messages, your images, your parameters, stays in the same shape, so switching between them is a one line change and you can compare results on your own data without rewriting integration code.

Switch with one string

Both models use the same endpoint and request shape. Change the model value from "google/gemini-3.1-flash-lite-preview" to "qwen/qwen2.5-vl-72b-instruct" and prompts, streaming and tools keep working.

Google: Gemini 3.1 Flash Lite Preview details & playground →Qwen: Qwen2.5 VL 72B Instruct details & playground →

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api14.com/api/v1",
  apiKey: process.env.API14_API_KEY,
});

const result = await client.chat.completions.create({
  "model": "qwen/qwen2.5-vl-72b-instruct",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

Google: Gemini 3.1 Flash Lite Preview vs Qwen: Qwen2.5 VL 72B Instruct

Specifications

Key differences

In depth

Switch with one string

More comparisons