NeuralSpark
NeuralSpark Get NeuralSpark

Providers · September 2026

Your provider. Your key.
Their invoice.

NeuralSpark runs on an API key from a provider you choose. You pay them for what you use, at their rates; NeuralSpark charges nothing for it. This page is the honest comparison, our pick, and the ten-minute setup.

Cheapest, with a trade

Meta Muse 1.3 Contributor

A strong all-rounder with images and search built in, at a fraction of the price — because Meta trains on what you send it. Fine for everyday work; not for anything confidential.

Costs about
US$0.10 in · US$0.20 out per million tokens
Standard tier
US$1.25 in · US$4.25 out, no training on your data
Search
Built in (US$2.50 per 1,000 searches)
Images
Muse Image, about 1¢ each
Get a Muse key ↗

NeuralSpark marks the Contributor model in Settings with "Meta trains on prompts and outputs" so you always know which one you're on. You can connect both and switch per conversation.

Prices

What each provider charges.

US dollars per million tokens, as published on each provider's own pricing page on 21 September 2026. A million tokens is roughly 700,000 words. Prices change; the links go to the source.

Provider · modelInputOutputSearchImagesYour data
DeepSeek · deepseek-flash$0.30 ($0.15 off-peak)$1.20 ($0.60 off-peak)Built in—Not used for trainingPricing ↗
DeepSeek · deepseek-v4-pro$1.32$3.96Built in—Not used for trainingPricing ↗
Meta · muse-spark-1.3-contributor$0.10$0.20$2.50 / 1k~1¢ eachUsed for trainingPricing ↗
Meta · muse-spark-1.3$1.25$4.25$2.50 / 1k~1¢ eachNot used for trainingPricing ↗
Google · gemini-3.5-flash-lite$0.30$2.505,000 free / month, then $14 / 1kNano Banana, ~4.5¢ eachFree tier may be used for training; paid tier is notPricing ↗
Google · gemini-3.8-flash$0.75$3.75as aboveas aboveas abovePricing ↗
OpenAI · gpt-5-mini$0.25$2.00$10 / 1kgpt-imageNot used for trainingPricing ↗
OpenAI · gpt-5.2$1.75$14.00$10 / 1kgpt-imageNot used for trainingPricing ↗
Anthropic · Haiku 4.5$1.00$5.00$10 / 1k—Not used for trainingPricing ↗
Anthropic · Sonnet 5$2.00$10.00$10 / 1k—Not used for trainingPricing ↗
xAI · grok-4.3$1.25$2.50Built ingrok-imagine, 4¢ eachSee xAI's termsPricing ↗

Each provider has its own page: NeuralSpark for DeepSeek · NeuralSpark for Meta Muse · NeuralSpark for Google Gemini · NeuralSpark for OpenAI · NeuralSpark for Anthropic Claude · NeuralSpark for xAI Grok.

"Your data" is our reading of each provider's published policy for API use; check theirs before sending anything sensitive. A chat subscription (ChatGPT Plus, Claude Pro, Gemini Advanced) is not an API key; you need a developer key from the pages linked.

Getting going

Ten minutes, once.

  1. Make a developer account

    DeepSeek: platform.deepseek.com. Meta: dev.meta.ai. Google: aistudio.google.com/apikey. OpenAI: platform.openai.com. Anthropic: platform.claude.com. xAI: console.x.ai.

  2. Add a little credit

    Most providers ask for prepaid credit — US$5 goes a long way on Flash or Muse. Set a monthly limit there if they offer one.

  3. Create a key

    Look for "API keys" and create one. Copy it once; the provider will not show it again.

  4. Paste it into NeuralSpark

    Settings › Connections › Models › Connect a provider. NeuralSpark recognises the provider from the key, lists the models it can reach, and tests the connection. Pick your default. The key is kept in Windows' protected storage and never shown again.

Settings › Connections › Models in NeuralSpark with a provider connected.

Questions

Before you choose.

How much will I actually spend?

A busy day of chat, a few Cowork tasks and some searching is usually a few hundred thousand tokens. On DeepSeek Flash that is a few cents; on Muse Contributor, less. Pictures are a cent or so each on Muse, about four cents on Google. Reasoning-heavy or long-document work costs more; the provider's dashboard shows it to the cent.

Why not just pick the cheapest?

The cheapest (Muse Contributor) is cheap because Meta trains on what you send. For a shopping comparison or a poster, that is fine. For client files, contracts or anything personal, use DeepSeek Flash or Muse Standard, which do not.

Can I connect more than one?

Yes. Connect several, choose a default, and switch per conversation from the composer. Chat's designer uses your default for the writing and whichever connected provider makes images.

Which one makes images?

Meta Muse, Google Gemini, OpenAI and xAI. DeepSeek and Anthropic are text-only; NeuralSpark labels them so, and Chat tells you which to connect if a design needs a picture.

Does NeuralSpark see my key or my usage?

No. The key lives in Windows' protected storage on your PC and goes only to the provider. Nothing about your usage is sent to us.