//Compare>Nous Portal vs Fireworks AI

Nous Portal vs Fireworks AI

Nous Portal gives you a pay-as-you-go API, models that are free only on Nous Portal and discounts on selected models. On Fireworks AI, Claude and GPT run on your own provider account.

Start with Nous PortalSet it up in Hermes Agent

Why Nous Portal

  1. 01

    Pay-as-you-go API, no subscription needed

    Nous Portal API keys bill per token from your credits, at the prices on the models page. No subscription is needed, and x402 (beta) needs no account.

  2. 02

    Models that are free only on Nous Portal

    Step 3.7 Flash is free on Nous Portal and a paid on-demand model on Fireworks AI. The Free filter on the models page lists every free model.

  3. 03

    Discounts on selected models

    Selected models sell below list price on Nous Portal, and some cost less than on Fireworks AI, such as GLM-5.3 Flash.

  4. 04

    One plan covers models, tools and hosting

    Paid plans include Hermes Agent usage, hosted tools and one Hermes Cloud agent. Setup is one command or one click in Hermes Desktop.

At a glance

A check marks where Nous Portal has the edge.

Topic
Nous Portal
Fireworks AI
Hermes Agent setup
One command, hermes setup --portal, or one click in Hermes Desktop.
A built-in Hermes Agent provider that needs a Fireworks AI API key.
Subscription plans
Edge: Plus, Super and Ultra include Hermes Agent usage, reset daily to monthly.
Fire Pass, its experimental coding pass, takes no new subscriptions.
Agent tools
Edge: Paid plans include web search, images, speech and a cloud browser.
MCP, web search and Python in its Responses API; search needs enabling.
Agent hosting
Edge: Every paid plan includes one always-on Hermes Cloud agent.
No hosted agents. Its docs point to AWS AgentCore Runtime.
Models
Edge: Claude, GPT, Gemini and Grok on launch day, beside open models.
Open models and Ember-1. Closed models run on your own provider account.
Free models
Edge: Some are free only here, such as Step 3.7 Flash. See the Free filter.
A starter credit. After it, accounts need a payment method.
How you pay
Pay-as-you-go API per token, with selected models below list price.
Prepaid per token or GPU time, with batch jobs at 50% of serverless.

Switch from Fireworks AI

  1. 01

    Run hermes setup --portal

    It signs Hermes Agent in to Nous Portal, lets you pick a model and turns on the Tool Gateway. Hermes Desktop does it in one click.

  2. 02

    Point your code at the Nous API

    Use an OpenAI-compatible client with a Portal API key at inference-api.nousresearch.com/v1.

  3. 03

    Keep your other providers

    Picking Nous Portal in hermes model keeps Fireworks AI configured, so /model can switch between them.

hermes setup --portal

Questions

Nous Portal or Fireworks AI: which is better for Hermes Agent?

Nous Portal gives you a pay-as-you-go API, models that are free only on Nous Portal and discounts on selected models. For Hermes Agent, paid plans add included usage, hosted tools and one Hermes Cloud agent, and setup is one command. The Hermes Agent docs call Nous Portal the recommended way to run Hermes Agent. Fireworks AI is also a built-in Hermes Agent provider, and it documents no hosted agents.

How does Nous Portal pricing compare with Fireworks AI?

Both bill from prepaid credits. Nous Portal sells selected models below list price: on Nov 9, 2026, GLM-5.3 Flash cost 20% less than on Fireworks AI serverless. Fireworks AI bills serverless per token, deployments by GPU time and training per token, with batch jobs at 50% of serverless rates and US-only models at 1.5x.

Does Nous Portal have free models?

Yes, and some are free only on Nous Portal: on Nov 9, 2026, Step 3.7 Flash was free on Nous Portal and paid on Fireworks AI, OpenRouter and Vercel AI Gateway. Beam 1.0 was also free on Nous Portal and paid on Fireworks AI. The Free filter on the models page shows the current list, and the Free plan covers free models with standard rate limits. Any account can also link an eligible ChatGPT account for supported inference.

Can I use Claude or GPT models through Fireworks AI?

Through Nexus Auto, harder turns go to an Anthropic or OpenAI model on your own account with that provider, billed there. Account admins connect that key with Provider Keys. Nous Portal lists Claude, GPT, Gemini and Grok models, including Claude Opus 5.5 and GPT-6 Astra, billed to your Nous Portal credits.

How do I switch from Fireworks AI to Nous Portal?

In Hermes Agent, run hermes setup --portal. It signs you in to Nous Portal, lets you pick a model and turns on the Tool Gateway, and Hermes Desktop does the same in one click. On an existing install, picking Nous Portal in hermes model keeps your other providers configured, so /model can switch between them. In your own code, point an OpenAI-compatible client at https://inference-api.nousresearch.com/v1 with a Portal API key, which you create after adding credits or starting a plan.

Can I host Hermes Agent on Nous Portal?

Yes, every paid Nous Portal plan includes one Hermes Cloud agent, up to the plan's included size, and larger or extra agents bill daily from credits. Without a plan, a small credit balance lets you deploy one. Fireworks AI documents no hosted agents, and its docs point to AWS AgentCore Runtime for deploying agents.

Is Fireworks AI open source?

Fireworks AI publishes the Nexus CLI under the Apache 2.0 license and a public cookbook on GitHub, and it serves open models. Fireworks AI documents no public source code for its serving platform. Nous Portal is a hosted account service. Hermes Agent, the agent Nous Portal is built around, is open source under the MIT license.

In detail

The Nous Portal API is pay-as-you-go

Nous Portal API keys bill per token from a credit balance, at the prices on the models page, and no subscription is needed for the API. API keys always use your credits, not a plan's included usage. Top-ups are paid by card through Stripe. For API calls, x402 (beta) lets you pay per request in Solana USDC with no account, API key or prepaid credits.

Fireworks AI also bills prepaid credits: per token for serverless, by GPU usage time for on-demand deployments and per training token for managed training. Batch jobs cost 50% of serverless rates, and US-only serverless models cost 1.5x the base rate. When credits run out with Auto Reload off, usage pauses, and contracted customers can move to post-paid invoicing.

Some models are free only on Nous Portal

Nous Portal offers free models, and the Free filter on the models page shows the current list. On Nov 9, 2026, Step 3.7 Flash, Beam 1.0, LongCat 2.5 and Solar Mini 4 were free on Nous Portal. Fireworks AI's model library listed Step 3.7 Flash and Beam 1.0 as paid on-demand models and did not list the other two. The Free plan covers free models with standard rate limits.

Fireworks AI's docs mention a starter credit. Once it runs out, an account without a payment method is suspended until one is added.

Selected models cost less than list price

Nous Portal lists hundreds of models, with new frontier models the day they launch, including Claude, GPT, Gemini, Grok, DeepSeek, Qwen, Kimi, GLM and MiniMax, with prices per million tokens. Selected models sell below their list price, and the Discounted filter on the models page shows the current list.

On Nov 9, 2026, GLM-5.3 Flash cost 20% less on Nous Portal than on Fireworks AI serverless. Fireworks AI serves open models and its own Ember-1. Through Nexus Auto, closed models from Anthropic and OpenAI run on your own account with that provider and bill there.

One plan covers Hermes Agent usage

Nous Portal is one account for Hermes Agent, the Nous inference API and Hermes Cloud. The Hermes Agent docs call Nous Portal the recommended way to run Hermes Agent. One command, hermes setup --portal, signs Hermes Agent in, lets you pick a model and turns on the Tool Gateway. Hermes Desktop does the same in one click.

Plus, Super and Ultra include Hermes Agent usage, within limits that reset daily, weekly and monthly. You can use any model, and pricier models use your included usage faster. If you use it up, you can switch to a free model, wait for the reset or continue with pay-as-you-go credits. Hermes Business gives a team a shared Hermes Agent deployment with one balance and per-member caps.

Fire Pass, Fireworks AI's experimental pass for personal, non-production coding, is not open to new subscriptions. Hermes Agent's docs list Fireworks AI as a built-in provider that needs only a Fireworks AI API key.

Paid plans include hosted agent tools

On paid plans, the Tool Gateway gives Hermes Agent web search, page extraction, image generation, text-to-speech, speech-to-text and a cloud browser, with no separate vendor accounts or keys. Hermes Agent can switch any single tool to your own vendor key whenever you want.

Fireworks AI's Responses API runs MCP, web search and Python tools server-side, and web search must be enabled for your account. The same API can store conversation state and run long jobs in the background.

Every paid plan includes a Hermes Cloud agent

Every paid Nous Portal plan includes one Hermes Cloud agent, up to the plan's included size: up to 4 GB RAM and 25 GB disk on Plus, 4 GB RAM and 100 GB disk on Super, and 8 GB RAM and 100 GB disk on Ultra. Larger or extra agents bill daily from credits.

Hermes Cloud runs Hermes Agent around the clock in the nearest available region. Each agent gets its own hardened container, and its files, memory and configuration persist whether it is running or stopped.

Fireworks AI documents no hosted agents. Its agent framework docs point to AWS AgentCore Runtime for deploying agents built on Fireworks AI models.

Sources (18), checked November 9, 2026

Every statement about Fireworks AI comes from the pages below. If something here is out of date, email support@nousresearch.com and we will correct it. Statements about Nous Portal come from this site and the Hermes Agent documentation.

More comparisons

  • Nous Portal vs OpenRouterModel router

    Nous Portal gives you a pay-as-you-go API, models that are free only on Nous Portal and discounts on selected models, plus hosted tools and Hermes Cloud.

  • Nous Portal vs Vercel AI GatewayAI gateway

    Nous Portal gives you a pay-as-you-go API, models that are free only on Nous Portal and discounts on selected models, plus hosted tools and Hermes Cloud.

  • Nous Portal vs Together AIInference provider

    Nous Portal gives you a pay-as-you-go API, models that are free only on Nous Portal and discounts on selected models, plus Claude, GPT and Hermes Cloud.