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.

Nemotron 3.5 Super VL vs Beam 1.0 501B A23B: API price, context & features · api14

NVIDIA: Nemotron 3.5 Super VL vs Reflection: Beam 1.0 501B A23B

A side-by-side comparison of NVIDIA: Nemotron 3.5 Super VL ($0.52 / 1M tokens) and Reflection: Beam 1.0 501B A23B ($1.05 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.

Specifications

	NVIDIA: Nemotron 3.5 Super VL	Reflection: Beam 1.0 501B A23B
Provider	NVIDIA	Reflection
Input	$0.52 / 1M tokens	$1.05 / 1M tokens
Output	$1.75 / 1M tokens	$3.50 / 1M tokens
Cache read	$0.07 / 1M tokens	$0.14 / 1M tokens
Context window	1,048,576	1,048,576
Max output	943,718	943,718
Input	text, image, video	text
Output	text	text
Tool calling
Structured output
Reasoning control
Added	Oct 15, 2026	Oct 16, 2026

Key differences

NVIDIA: Nemotron 3.5 Super VL is 50% cheaper on input tokens.
NVIDIA: Nemotron 3.5 Super VL is 50% cheaper on output tokens.
A request with 1,000 input and 500 output tokens costs $0.00140 with NVIDIA: Nemotron 3.5 Super VL and $0.00280 with Reflection: Beam 1.0 501B A23B.

In depth

NVIDIA: Nemotron 3.5 Super VL is made by NVIDIA, and Reflection: Beam 1.0 501B A23B is made by Reflection. Both are text models in the sense that their output is text, and both are served through api14 alongside each other. The two providers are different companies, so the models come from separate training efforts even though they sit behind the same API surface here.

The clearest split between them is in input modalities. Nemotron 3.5 Super VL accepts text, images and video as input, while Beam 1.0 501B A23B accepts text only. Both list a context length of 1,048,576 tokens, so there is no meaningful difference in how much you can feed them. On parameters, the two lists are nearly identical: both support frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs and top_p. Beam 1.0 501B A23B adds reasoning_effort on top of that set, which gives you a knob to dial reasoning intensity that Nemotron 3.5 Super VL does not expose. Nemotron 3.5 Super VL is the cheaper of the two on both input and output, so if cost per token matters, that is the one to lean toward.

For practical selection, start with the input you actually have. If your task involves screenshots, scanned documents, diagrams, photos or video frames, Nemotron 3.5 Super VL is the only one of the two that can take that material directly, so it is the default choice. If your input is plain text, both are viable, and the decision comes down to cost and control. Nemotron 3.5 Super VL is cheaper on both input and output, which makes it attractive for high volume text work such as classification, extraction, summarisation and bulk generation. Beam 1.0 501B A23B is the one to consider when you want explicit control over how much reasoning the model does, since reasoning_effort is available there and not on the NVIDIA model. Otherwise the two are close enough in context length and supported parameters that either can handle long documents, tool calls and structured output.

Calling either model through api14 works the same way, because the gateway exposes an OpenAI-compatible API. You point your client at the same base URL, send the same chat completions style request body, and change only the model id. Use "nvidia/nemotron-3.5-super-vl" when you want the NVIDIA model, and "reflection/beam-1.0-501b-a23b" when you want the Reflection model. Everything else in the request, including messages, tools, response_format and sampling parameters, stays as it is, subject to the parameter lists above. That means you can switch between them for comparison or fallback by editing a single string, without rewriting your integration.

There is a case where Beam 1.0 501B A23B is still the better pick even though it costs more and takes text only. If your workload is text based and you specifically need the reasoning_effort control to tune how much internal reasoning the model performs, that parameter is only on the Reflection model. In that situation the extra cost and the lack of image and video input do not matter, because your inputs are text and your priority is steering reasoning depth rather than minimising spend. For everything else, especially anything involving images or video, or any text workload where the cheaper rates are the deciding factor, Nemotron 3.5 Super VL is the more natural default.

Switch with one string

Both models use the same endpoint and request shape. Change the model value from "nvidia/nemotron-3.5-super-vl" to "reflection/beam-1.0-501b-a23b" and prompts, streaming and tools keep working.

NVIDIA: Nemotron 3.5 Super VL details & playground →Reflection: Beam 1.0 501B A23B 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": "reflection/beam-1.0-501b-a23b",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

NVIDIA: Nemotron 3.5 Super VL vs Reflection: Beam 1.0 501B A23B

Specifications

Key differences

In depth

Switch with one string

More comparisons