Artificial Analysis Model Metrics API
Look up public Artificial Analysis model intelligence, median output speed and weighted Intelligence Index task cost by model key or exact name. Extended single and batch lookups include the source estimated flag, token prices and derived non-hallucination percentage from live model detail pages. Unreported measurements are null; residential HTTP serves the models page and detail pages without a browser.
Endpoint health
Loading verification…
Connect your agent
Connect through MCP so your agent can find this API, read its docs, and call its endpoints.
Add the server from your terminal.
claude mcp add --transport http agent-data https://agent-data.motie.dev/mcpIn Claude Code, run /mcp, select agent-data, then Authenticate. Sign in to your agent-data account in the browser.
Then ask your agent to use the Artificial Analysis Model Metrics API for your task.
Call it over HTTP
Get an API keySign up free, generate an API key, and set it as AGENT_DATA_API_KEY in your environment. Your first 100 calls each month are free.
curl --request GET --get 'https://3kpkg0jfvk.execute-api.us-east-2.amazonaws.com/prod/10600049-a8af-420f-80bb-93ee03e948dd/v1/model-metrics' \
--header "Authorization: Bearer $AGENT_DATA_API_KEY"Endpoints
Inputs and response fields for every endpoint in this API.
GETModel metrics/v1/model-metrics
Fetch live model data; unknown keys/names return JSON 404, ambiguous names JSON 422.
Input
model_keyOptionalQuery · string | nullExact source model slug, e.g. gemini-3-5-flash-lite. Supply one of model_key or model_name.
Example: gemini-3-5-flash-lite
model_nameOptionalQuery · string | nullExact Artificial Analysis model name, case-insensitive. Supply one of model_key or model_name.
Example: Gemini 3.5 Flash-Lite
Response
model_keystringtransportstringmodel_namestringsource_urlstringcost_per_task_usdnumber | nullintelligence_index_scorenumber | nulloutput_tokens_per_secondnumber | nullGETModel metrics extended/v1/model-metrics/extended
Fetch live detail metrics from the selected model page, with null for unavailable measurements.
Input
model_keyOptionalQuery · string | nullExact source model slug. Supply exactly one of model_key or model_name.
Example: mimo-v2-6-flash
model_nameOptionalQuery · string | nullExact Artificial Analysis model name (case-insensitive). Supply exactly one of model_key or model_name.
Example: MiMo-V2.6-Flash
Response
model_keystringtransportstringmodel_namestringsource_urlstringcache_hit_pricenumber | nullcost_per_task_usdnumber | nullprice_1m_input_tokensnumber | nullprice_1m_output_tokensnumber | nullintelligence_index_scorenumber | nulloutput_tokens_per_secondnumber | nullnon_hallucination_rate_pctnumber | nullomniscience_hallucination_ratenumber | nullintelligence_index_is_estimatedboolean | nullGETModel metrics extended batch/v1/model-metrics/extended/batch
Look up model detail pages concurrently; per-item errors do not fail other models.
Input
model_keysOptionalQuery · string | nullComma-separated exact source slugs, 1 to 20. Supply exactly one of model_keys or model_names.
Example: mimo-v2-6-flash,step-5
model_namesOptionalQuery · string | nullComma-separated exact model names, case-insensitive, 1 to 20; CSV-quote any name containing a comma. Supply exactly one of model_keys or model_names.
Example: MiMo-V2.6-Flash,Step 5 Preview
Response
itemsobject[]items[].errorstring | nullitems[].inputstringitems[].metricsobjectitems[].metrics.model_keystringitems[].metrics.transportstringitems[].metrics.model_namestringitems[].metrics.source_urlstringitems[].metrics.cache_hit_pricenumber | nullitems[].metrics.cost_per_task_usdnumber | nullitems[].metrics.price_1m_input_tokensnumber | nullitems[].metrics.price_1m_output_tokensnumber | nullitems[].metrics.intelligence_index_scorenumber | nullitems[].metrics.output_tokens_per_secondnumber | nullitems[].metrics.non_hallucination_rate_pctnumber | nullitems[].metrics.omniscience_hallucination_ratenumber | nullitems[].metrics.intelligence_index_is_estimatedboolean | nulltransport?string, default: curl_cffi_residentialFields marked ? are optional.
Don’t see the endpoint you need?
Ask agent-data to add an endpoint or return more fields. Extensions are free and keep existing endpoints working.
An account is required to submit an extension. Prepare your request, then sign in to build.