Skip to main content
GET
Model ranking
Pass category=all, a comma-separated list, or repeated category parameters to rank several tasks in one call. One category returns a flat object (with pick). Two or more return picks and categories. Set includeItems=false to drop per-model items; frontier still adds frontier_rank. pick is null when nothing scored for that task.

Authorizations

Authorization
string
header
required

Enter your API key (starts with 'ek-')

Query Parameters

category
string
default:chat

One of chat, coding, reasoning, math, science, creative, vision, long_context, agentic, factuality, safety, roleplay, summarization, translation; a comma-separated subset; repeated category parameters; or all. Case is ignored. all cannot be combined with other categories. One category returns a flat object. Two or more return picks and categories.

costTier
enum<string>
default:medium
Available options:
low,
medium,
high,
xhigh,
max
difficulty
enum<string>
default:medium
Available options:
trivial,
easy,
medium,
hard,
extreme
inputTokens
integer
default:2000
Required range: x >= 0
maxOutputTokens
integer
Required range: x >= 1
endpoint
enum<string>
default:chat
Available options:
chat,
messages,
responses
domain
string

Professional domain, e.g. legal, medical, finance.

frontier
number

When set, each category also includes frontier_rank (Pareto mix of quality vs price).

Required range: 0 <= x <= 1
includeItems
boolean
default:true

Set false to omit per-model items. frontier still adds frontier_rank. Leave true for the Cost vs quality chart.

Response

Success

The response is of type object.