Models

The right model for every agent, every job.

Browse a curated catalog of leading language models from Anthropic, OpenAI, Google, and more. Compare capabilities, modalities, and cost — and give every agent exactly the foundation it needs.

Start for free Request a demo

Model catalog · 200+ modelsalways current

claude · sonnet

anthropic · text · image · file

gpt · omni

openai · text · image · audio

gemini · pro

google · text · image · video

claude · haiku

anthropic · text

model configpreset

agentbuilt on

agent templatestandardized

The problem

Most platforms pick the model for you.

Model choice is an operational decision — capability, modality, and cost all shift the outcome of the work your agents do every day.

A default you cannot change is a decision made without you.

One size fits no one

A single default model cannot serve research agents, multimodal content tools, and high-volume support automations equally well. Cost, quality, and capability trade-offs matter — and they compound.

No visibility into what is running where

When a model is deprecated or needs to change, teams often cannot tell which agents depend on it until something breaks.

No control over what gets used

Without org-level governance, anyone can deploy agents on any model, regardless of cost, compliance requirements, or internal standards.

What is a Model

Every leading model. One place to compare and choose.

Opal's model catalog gives your organization a single place to discover, evaluate, and govern the models that power your digital workforce.

01
Curated catalog
Browse over 200 models from leading providers including Anthropic, OpenAI, and Google — organized, searchable, and always current as Opal monitors for new releases and deprecations.
02
Filter by modality
Find the right model by filtering on what it can take in and produce: text, images, audio, video, or files. Match each agent's task to a model built for it.
03
One AI budget, at cost
Every model runs on Opal credits, so your entire AI budget lives in one place instead of a dozen separate subscriptions. Model usage is passed through at cost, and credit pricing simply reflects what the providers charge.
04
Automatic optimization
Token caching and other cost optimizations are applied automatically — without configuration. Lower costs, no extra work.
05
Linked resources
See every Model Config, Agent, and Agent Template that depends on a given model. Before changing or disabling a model, know exactly what is affected.
06
Org-level governance
Administrators can disable models that do not meet internal standards. The catalog your team sees only shows what has been approved for use.

Model · catalogorg approved

claude · sonneton

gpt · omnion

gemini · proon

legacy · v2off

agent · researchgemini · pro

agent · triagegpt · omni

agent · draftingclaude · sonnet

different models per agent

How it works

Browse, compare, decide, govern.

Step 01

Open the catalog

Browse all available models or search by name. Filter by provider or by the input and output modalities a model supports.

Step 02

Review the details

Open any model to see its specifications, capabilities, supported parameters, and token cost in Opal credits.

Step 03

Check dependencies

Use the Linked Resources view to see which Model Configs, Agents, and Agent Templates rely on a given model — essential context before making any change.

Step 04

Select and build

Use a model directly to create an Agent or Agent Template, with the right foundation in place from the start.

Step 05

Govern the catalog

Administrators disable models that fall outside policy. Every member who opens the catalog only sees what the organization has approved.

Your AI budget

Manage your entire AI budget in one place.

Opal is not in the business of making money on tokens. Model usage is passed through at cost through the credit system, so consolidating your AI spend here is simply the cheaper way to run it.

No stacked subscriptions

You do not need separate seats and plans with Anthropic, OpenAI, Google, and everyone else. One agreement with Opal covers every model in the catalog.

Model usage at cost

Credits are priced to reflect what providers charge for tokens. We do not mark up model usage, and we do not make margin on your consumption.

One pool, every workload

Agents, tasks, flows, and threads all draw from the same credit balance, so a single budget covers everything your organization runs.

Spend you can actually see

Because every workload settles in the same unit, you can compare the real cost of models, teams, and workflows without reconciling separate invoices.

Key benefits

Better choices. Lower costs. Full control.

Always-current catalog

Opal continuously monitors for new and deprecated models. You always work from a live, supported list — not yesterday's state.

Confident selection

Compare capabilities, modalities, and cost side by side before committing to a model. No guesswork required.

Pricing you can actually compare

Token costs shown in Opal credits make it straightforward to understand and compare what running each model actually costs.

Cost savings without configuration

Token caching reduces credit usage automatically. No manual optimization required.

Governance without friction

Administrators maintain control over which models are in use. Traceability through Linked Resources makes change management predictable and informed.

Automatic data protection

Zero data retention, enabled automatically.

When you use a model and provider that supports zero data retention, Opal enables it automatically. No extra configuration required. Eligible workloads benefit from stronger data handling without any additional steps on your end.

See it in practice

Matching the model to the work.

Capability, modality, and cost — weighed once, up front, with full visibility into what depends on the choice.

Case 01

Support Ops

High-volume support agent

A support operations lead is building an agent for triaging and drafting replies to inbound tickets. She opens the model catalog, filters by text input and output, and compares two candidate models on capability and cost per credit. She chooses the model that balances quality against a lower credit cost — and because the model supports zero data retention, Opal enables it automatically.

Filter: Text input · text output
Decision: Quality against credit cost
Outcome: ZDR enabled automatically

Case 02

Marketing

Multimodal marketing content

A marketing manager needs an agent that produces both campaign copy and supporting visuals. He filters the catalog by text and image output to surface only models that can deliver both, reviews specifications and cost, and selects the right option for his creative and budget requirements.

Filter: Text and image output
Decision: Creative and budget fit
Outcome: One agent, copy and visuals

Case 03

Administration

Governance and impact review

An administrator reviews the catalog and disables models that do not meet internal policy. Before disabling one, she opens its Linked Resources view and checks which agents and templates currently depend on it — understanding the impact before coordinating the change with her team.

Action: Disable out-of-policy models
Check: Linked Resources view
Outcome: Impact understood before change

Security and compliance

Model choice that stays governed and traceable.

Role-Based Access Control

Administrators control which models are available across the organization.

Zero Data Retention

Enabled automatically for models and providers that support it — no configuration required.

Linked Resources Traceability

Every Model Config, Agent, and Agent Template tied to a model is visible before any change is made, so governance is informed by real usage.

Audit Log

Model-level changes and administrative actions are tracked in the platform's activity log.

Certifications

ISO 27001, ISO 42001, and SOC 2 Type II certifications are underway.

Part of a complete platform

Models connect across Opal.

01AgentsEvery agent is built on a model selected from the catalog — and Agent Templates standardize that choice for repeatable deployment.02GovernanceAdministrators control which models are available through org-level settings.03GuardrailsGuardrails apply input and output enforcement regardless of which model an agent uses.04ConnectorsAgents use models alongside Connectors to reason over and act on live data.