Artificial Analysis

Models

Third-party benchmark charts for every model Artificial Analysis tracks, with the ones sharedcode has enabled marked on each chart.

—Models tracked

—sharedcode enabled

—Index version

—API tier

—Requests left

—Source

sharedcode enabled enabled via an approximate match (≈) not enabled in sharedcode running locally (sharedcode serves it here)

Intelligence Index vs. Cost per Intelligence Index Task

Artificial Analysis Intelligence Index · Weighted average cost (USD) per Artificial Analysis Intelligence Index task

Loading…

Why these models?

Local router

—

LLM stats only. Other kinds show whether they are up and how many calls they took.

—Servers

—LLM replicas up

—LLM running

—LLM waiting

—LLM cap

Routing

No models configured

Operator list in vllm-router.yaml. Each model carries its own fallback hops — there is no separate external-provider panel.

vLLM serving

—

LLM running now, request history, last-turn stats, and process logs. Other kinds: up/down and calls.

—LLM running now

—LLM queued

—LLM cap

—LLM replicas up

—User unit

LLM requests

LLM replicas

live from /metrics

Server	Pool	Replica	Running	Waiting	KV cache	tok/s	Cap	Status
No replicas reported

Decision models

status and calls

Pool	Kind	Status	Calls
No decision models reported

Other models

status and calls

Pool	Kind	Status	Calls
No other models reported

Last turns

from process logs

LLM routed totals

vllm_mux_route_total · LLM only

Process logs

last 250 lines

Models

Models

Artificial Analysis API key required

Local models

Models on the Pareto chart

Filters

Chart settings

Intelligence Index vs. Cost per Intelligence Index Task

Intelligence

Speed

Cost per Task

—

Routing

—

LLM requests

LLM replicas

Decision models

Other models

Last turns

LLM routed totals

Process logs