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.

Inception: Mercury 2 vs Qwen: Qwen3 VL 8B Instruct

A side-by-side comparison of Inception: Mercury 2 ($0.88 / 1M tokens) and Qwen: Qwen3 VL 8B Instruct ($0.88 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.

Specifications

	Inception: Mercury 2	Qwen: Qwen3 VL 8B Instruct
Provider	Inception	Qwen
Input	$0.88 / 1M tokens	$0.88 / 1M tokens
Output	$2.63 / 1M tokens	$2.63 / 1M tokens
Cache read	$0.09 / 1M tokens	$0.42 / 1M tokens
Context window	128,000	262,144
Max output	50,000	235,929
Input	text	image, text
Output	text	text
Tool calling
Structured output
Reasoning control
Added	Mar 4, 2026	Oct 14, 2025

Key differences

Both have about the same input price.
Qwen: Qwen3 VL 8B Instruct reads 2.0× more context (262,144 tokens).
Only Inception: Mercury 2 supports reasoning effort control.
A request with 1,000 input and 500 output tokens costs $0.00219 with Inception: Mercury 2 and $0.00219 with Qwen: Qwen3 VL 8B Instruct.

In depth

Inception: Mercury 2 is a text model from Inception, released on 4 March 2026. Qwen: Qwen3 VL 8B Instruct is a text model from Qwen, released on 14 October 2025. Both are served through api14 under the text category, so they occupy the same slot in your integration even though they come from different providers and were built at different times.

The clearest split is in context and modalities. Qwen3 VL 8B Instruct accepts both image and text as input and returns text, while Mercury 2 accepts text only and returns text. Qwen3 VL 8B Instruct also has the larger context window at 262144 tokens against 128000 for Mercury 2, so it can hold roughly twice as much material in a single call. On reasoning, Mercury 2 is the one with dedicated reasoning support: its parameter list includes include_reasoning, reasoning and reasoning_effort, none of which appear for Qwen3 VL 8B Instruct. The two are about the same price on both input and output, so cost is not a deciding factor here.

For tasks that involve screenshots, diagrams, scanned pages or any other image alongside text, Qwen3 VL 8B Instruct is the only option of the two, since Mercury 2 cannot take image input at all. The same applies when you need to fit a very large document, transcript or codebase into one prompt, where the larger context window of Qwen3 VL 8B Instruct gives you more room before you have to chunk. Mercury 2 is the better fit when the task benefits from explicit reasoning control, because you can ask for reasoning output and tune reasoning effort, and when the input is purely text and comfortably within 128000 tokens. Qwen3 VL 8B Instruct also exposes a broader set of sampling and decoding parameters, including frequency_penalty, presence_penalty, repetition_penalty, logit_bias, seed, top_k, top_p, logprobs and top_logprobs, which matters if you rely on seeded reproducibility or token level inspection. Mercury 2 counters with response_format, structured_outputs, tools and tool_choice, all of which Qwen3 VL 8B Instruct also supports, so structured output and tool calling are available on either side.

Calling either model through api14 works the same way, because the API is OpenAI compatible. You point your client at the api14 base URL, send your request to the chat completions endpoint with your api14 key, and set the model field to the id you want. Switching between them is a one line change: use "inception/mercury-2" for Mercury 2 and "qwen/qwen3-vl-8b-instruct" for Qwen3 VL 8B Instruct. The message format, the tools and response_format fields and the streaming behaviour stay the same, so you can keep one code path and swap the id, or route between the two at runtime based on whether the request carries an image.

There is one case where the reasoning model is still the better pick even though the other model is more flexible overall. If your workload is text only, fits inside 128000 tokens, and you want the model to work through a problem step by step with control over how much reasoning it does, Mercury 2 gives you that through include_reasoning and reasoning_effort, and Qwen3 VL 8B Instruct does not offer those parameters. In that situation the extra context and image support of Qwen3 VL 8B Instruct buy you nothing, and Mercury 2 is the more direct choice.

Switch with one string

Both models use the same endpoint and request shape. Change the model value from "inception/mercury-2" to "qwen/qwen3-vl-8b-instruct" and prompts, streaming and tools keep working.

Inception: Mercury 2 details & playground →Qwen: Qwen3 VL 8B 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/qwen3-vl-8b-instruct",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

More comparisons

Mercury 2 vs Qwen3 VL 8B Instruct: API price, context & features · api14