Which is cheaper, Qwen3 VL 32B Instruct or Qwen2.5 72B Instruct?

Qwen3 VL 32B Instruct is the cheaper of the two: one 1,000-word request costs about 196 toman, about 1.5× cheaper than Qwen2.5 72B Instruct.

Which one has the larger context window?

Qwen3 VL 32B Instruct, with 131,072 tokens.

Qwen3 VL 32B Instruct vs Qwen2.5 72B Instruct

Both sit behind the same key and the same code on uttapen — the only thing that changes is the model string, so you can run each of them against your own workload without touching anything else. On cost, Qwen3 VL 32B Instruct comes out about 1.5× cheaper.

Feature	Qwen3 VL 32B Instruct	Qwen2.5 72B Instruct
Provider	Qwen (Alibaba)	Qwen (Alibaba)
Model id	qwen/qwen3-vl-32b-instruct	qwen/qwen-2.5-72b-instruct
Context	131,072 tokens	32,768 tokens
Max output	32,768 tokens	16,384 tokens
Input / 1M tokens	30,173 toman	104,445 toman
Output / 1M tokens	120,692 toman	116,050 toman
≈ one 1,000-word request	196 toman	287 toman
Input caching	No	No
Image input	Yes	No
File input	No	No
Tool calling	Yes	Yes
JSON output	Yes	Yes
Reasoning mode	No	No

Which one for what?

Qwen3 VL 32B Instruct

Reading screenshots, invoices, and scanned forms
Agents that call your own APIs and database
High-volume work where cost per call decides

196 toman per 1,000 words · Model page

Qwen2.5 72B Instruct

Agents that call your own APIs and database
High-volume work where cost per call decides

287 toman per 1,000 words · Model page

Run both with the same code

Swap the model value between the two ids and send the same request twice; what each answer cost comes back in the X-Uttapen-Cost-Toman header.

curl https://api.uttapen.ir/v1/chat/completions \
  -H "Authorization: Bearer sk-up-…" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3-vl-32b-instruct",
    "messages": [{"role": "user", "content": "Introduce yourself in one sentence."}],
    "stream": true
  }'

from openai import OpenAI

client = OpenAI(
    base_url="https://api.uttapen.ir/v1",
    api_key="sk-up-…",
)

stream = client.chat.completions.create(
    model="qwen/qwen3-vl-32b-instruct",
    messages=[{"role": "user", "content": "Introduce yourself in one sentence."}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="", flush=True)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.uttapen.ir/v1",
  apiKey: "sk-up-…",
});

const stream = await client.chat.completions.create({
  model: "qwen/qwen3-vl-32b-instruct",
  messages: [{ role: "user", content: "Introduce yourself in one sentence." }],
  stream: true,
});
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

<?php
// composer require openai-php/client guzzlehttp/guzzle
$client = OpenAI::factory()
    ->withBaseUri('https://api.uttapen.ir/v1')
    ->withApiKey('sk-up-…')
    ->make();

$result = $client->chat()->create([
    'model' => 'qwen/qwen3-vl-32b-instruct',
    'messages' => [['role' => 'user', 'content' => 'Introduce yourself in one sentence.']],
]);
echo $result->choices[0]->message->content;

package main

import (
	"context"
	"fmt"

	"github.com/openai/openai-go"
	"github.com/openai/openai-go/option"
)

func main() {
	client := openai.NewClient(
		option.WithBaseURL("https://api.uttapen.ir/v1"),
		option.WithAPIKey("sk-up-…"),
	)
	resp, err := client.Chat.Completions.New(context.Background(), openai.ChatCompletionNewParams{
		Model:    "qwen/qwen3-vl-32b-instruct",
		Messages: []openai.ChatCompletionMessageParamUnion{openai.UserMessage("Introduce yourself in one sentence.")},
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(resp.Choices[0].Message.Content)
}

model = "qwen/qwen-2.5-72b-instruct"

More comparisons: all pairs · model rankings · full catalogue

base_url = https://api.uttapen.ir/v1