Decision system · live data

Choose an AI model with evidence, not vibes.

Explore the market, inspect the source trail, shortlist the contenders, then test the decision against your workload and monthly budget.

Model record Loading… Reading models.json

Workspace Guest mode Saved on this device

Decision trail 4 evidence labels Facts stay separate from judgment

Read the label before the number

Pricing and specifications are not the same kind of evidence as a provider benchmark or a benchr recommendation.

OfficialProvider documentation
ProviderProvider-published eval
BenchrVerified record
EditorialRating or decision logic

Guest · local

Loading the current model record…

Explore

Narrow the field

Editorial

Capability order uses benchr ratings and provider results. It is a decision aid, not a lab leaderboard.

Live inventory

Models in view

All providers · all tiers

Start here

Browse the verified model record

Benchr

The interactive inventory reads from the same model data that powers benchr rankings and calculators.

Open the static rankings

No JavaScript?

The research still works

Official

Use the pricing index, comparison tool, and lifecycle tracker while this workspace loads.

Browse verified pricing

This workspace needs JavaScript for filtering and local saves.

The rest of benchr remains available: use model rankings, comparison, or verified pricing.

Decision workspace

Turn a shortlist into a buying decision

Set one workload, its monthly token volume, and the budget ceiling. The result explains the trade-off and keeps the assumptions visible.

Guest workspace Local to this browser

Scenario

Describe the workload

Primary workload

Input tokens / month

million

Output tokens / month

million

Monthly API budget

USD

Decision lens Balanced Quality first Cost first

Rank my shortlist only

Estimate boundary

Token prices only. Cache, batch, tools, taxes, and self-hosting infrastructure are excluded.

Recommendation

Your ranked fit

Run the scenario to rank the market.

You will get three candidates, estimated monthly cost, and a plain-English reason for each position.

Shortlist compare

Keep the trade-offs on one screen

No models shortlisted yet. Add candidates from Explore or from a recommendation.

Guest library

Save the decision on this device

No benchr account is created. Your watchlist, collections, and workspaces use browser storage. Export a backup before clearing browser data or changing devices.

Watchlist

Watch a model to surface matching lifecycle and data changes.

Collections

Freeze the current shortlist as a named collection.

Saved workspaces

Save filters, shortlist, and the current decision scenario together.

Portable backup

Move guest data with JSON

Export includes local workspaces, collections, and watchlist. Import merges a valid benchr guest backup into this browser.

Import & merge

Internal alerts

Price, release, and retirement signals from the record

These alerts are generated inside this page from benchr history and deprecation data. This guest version does not send email, browser push, or account notifications.

Only changes matching my watchlist

Loading record

Benchr

Reading lifecycle and price history…

The latest verified entries will appear here.

Method, not magic

How the decision workspace works

The useful part is not a single winning score. It is the trail from an official price to a workload assumption, an editorial fit score, and a decision you can revisit when the market changes.

01
Discover
Search the complete tool-layer model list by name, provider, API ID, tier, license, price, or use case.
02
Inspect
Open a model panel to separate official specifications, provider-published evaluations, benchr records, and editorial judgments.
03
Decide
Rank the field against one workload, visible capability weights, token volume, and a monthly API budget.
04
Keep the trail
Save the workspace locally, watch the chosen models, export a backup, or share the current state in a URL.

FAQ

Before you save a workspace

Does the benchr workspace create an online account?

No. This version is guest-only. Watchlists, collections, and saved workspaces are stored in the browser on the current device. Export a JSON backup if you need to move them.

Are the recommendation scores official model benchmarks?

No. The decision score is an editorial calculation from the workload weights, available benchmark fields, benchr capability ratings, and your budget. Official facts, provider-published evaluations, and editorial estimates are labelled separately.

What does the monthly cost estimate include?

It multiplies the input and output token volumes you enter by the current per-million token prices in the benchr data file. It does not assume cache discounts, batch discounts, tool-call fees, taxes, or self-hosting infrastructure costs.

Can I share a model shortlist?

Yes. The Share button copies a URL carrying the current filters, shortlist, and decision scenario. Saved guest collections and other local library items are not included in the link.