ToolsCompaniesAbout

HUD

by HUDYC W25No ratings yet

A platform for creating, running, and reviewing reinforcement-learning tasks and benchmarks.

Visit hud.ai

Linked from hud.ai · checked 1 Nov

Tool type
RL task and benchmark platform
Made for
Researchers and task-building teams
Runs on
Web app and Python
The web app supports no-code task creation; the SDK defines environments in Python. The quickstart runs environments locally.
Last checked
1 Nov 2026
4 product sources

What it does

HUD is an RL task and benchmark platform for researchers, task authors, and teams evaluating or training models. Tasks combine an environment where an agent acts with a grader that evaluates the result. You can create tasks in the web app from files and a rubric, import Harbor tasks, or define environments in Python with the HUD SDK. Tasksets group tasks for evaluation and review.

Run tasks against one or more models, with configurable repetitions. HUD records each attempt as a trace containing the agent’s trajectory and reward; traces can be replayed and reviewed. QA agents examine tasks and traces for issues such as reward hacking, grader mistakes, and gaps between prompts and criteria. Teams can revise tasks, move them through review, publish benchmarks, or train models on tasksets.

The web workflow requires a HUD account and credits for runs and QA; a new team starts without credits. SDK users need Python and a HUD API key, and can run an environment locally with the CLI before viewing traces on hud.ai. Usage draws on credits for environment runtime and model calls; model calls through HUD’s gateway include a platform fee, while using your own provider key also incurs charges from that provider. Evaluation results depend on the task, grader, and model; QA reviews are billed separately.

What it can do

  • Create tasks from files and rubrics in the web app
  • Define environments and tasks in Python with the SDK
  • Import Harbor tasks and runs
  • Run tasksets against models with configurable repetitions
  • Record and replay traces with trajectories and rewards
  • Review tasks and traces with QA agents

Works with

  • Harbor task imports
  • Model providers through HUD's gateway or user-supplied provider keys
  • REST API

Before you choose

  • A new team starts with no credits; environment runtime and model use consume credits.
  • Model calls through HUD's gateway are charged by token use plus a 5% platform fee; using a provider key also means the provider bills for tokens.
  • QA reviews incur model-call and environment-runtime costs, so reviewing traces adds usage.

Listed under

Get started

Sign in at hud.ai; runs require credits. SDK use requires an API key.

  1. Create a HUD account and, for web runs, add credits or subscribe to Pro.
  2. For SDK use, install the HUD CLI or Python package and obtain an API key.
  3. Create or import a task, run it against a model, then inspect its trace and optional QA review.
Create a HUD account

Plans and prices

Prices in USD

Prepaid USD credits cover usage; Pro costs $50/month and adds $50 credits. Compute charges depend on size/resources: EC2 has a five-minute minimum then per-minute billing; Modal uses per-second CPU/memory/GPU billing. Gateway model tokens add a 5% platform fee. With your own provider key, HUD charges 5% of the equivalent call cost and the provider separately bills tokens. QA also consumes model and runtime usage.

Pay as you goUsage-basedAdd credits; runs consume environment runtime and model usage.
  • No subscription
  • Usage is paid from credits added to the team's balance
Pro$50/moMonthly subscription; each paid month adds $50 in credits.
  • Each paid month adds $50 of credits
  • Higher limits than Pay as you go
EnterpriseCustomContact sales.
  • Custom pricing
  • No credit check before work starts
  • Runs up to 24 hours
  • Unlimited storage
  • No per-minute limit on model calls
  • Sign in with HUD for a customer's own product users

Source: docs.hud.ai/platform/billing.md, checked 1 Nov 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 1 Nov 2026. Each source supports part of the overview; availability and prices can change.

Ratings and reviews

No ratings for HUD yet.

Be the first to rate it

What’s changed

Free account

See what’s changed at HUD

Launches, price moves and pivots, each with a date and a source. Free with an account.

Signals

Free account

See HUD’s signals

Launch traction, hiring and release activity, tracked over time. Free with an account.

Data and security

Free account

See how HUD handles your data

Training on your files, compliance, hosting and the models behind it. Free with an account.

Compare with similar tools

Free account

Compare HUD side by side

Prices, platforms and features against similar tools, with sources. Free with an account.

Where this comes from

How we research
Company website
hud.ai: product, pricing and policy pages
GitHub
Public code and releases
Web archive
When the site first appeared
YC directory mirror
Batch and company name
Web search
Confirms the official site
Social accounts
Accounts the company links, and its posts on X
Human review
Checked by a person
manual_review

Last checked 1 Nov 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.