The frontier, in one request.A free JSON API for agents and developers
Current models per lab, their SI Scores, prices and sources. Public JSON, no API key, no signup.
$ curl -fsSL https://superindex.app/api/v1/frontier.json
{
"schema": "superindex.frontier/1",
"generated_at": "2026-10-09T06:15:14.937Z",
"labs": [
{
"provider": "anthropic",
"provider_name": "Anthropic",
"frontier": [
{
"id": "anthropic/claude-fable-5-1",
"name": "Claude Fable 5.1",
"si_score": 80.2,
"rank": 1,
"price_in": 10,
"price_out": 50,
"context_window": 1000000,
"url": "https://superindex.app/models/anthropic/claude-fable-5-1/"
}
],
"latest_release": {
"id": "anthropic/claude-haiku-5-5",
"name": "Claude Haiku 5.5",
… Endpoints
- GET /api/v1/frontier.json Frontier by lab. A short list per lab: its top-ranked model, ranked models close behind it, and its newest flagship awaiting results. Schema
- GET /api/v1/models.json All text models. Every text model, ranked models first, then provisional releases by date. Schema
- GET /api/v1/index.json Endpoint index. Endpoints, schema IDs and schema URLs, for clients that discover the API.
Start here
Fetch the frontier file when an agent needs to discover current models. Fetch the full models file for comparisons or a local catalog. Requests are served as static Cloudflare Pages files: no Function invocation, credentials or request charge.
Keep unknowns unknown. A provisional model may be a new release awaiting public benchmark results. Null prices and scores are missing evidence, not zero cost or zero capability. Check generated_at before making a time-sensitive choice.
curl -fsSL https://superindex.app/api/v1/frontier.json const response = await fetch("https://superindex.app/api/v1/frontier.json");
if (!response.ok) throw new Error(`HTTP ${response.status}`);
const data = await response.json();
if (data.schema !== "superindex.frontier/1") throw new Error("Unsupported schema");
for (const lab of data.labs) {
console.log(lab.provider_name, lab.frontier.map(model => model.id));
} import json
from urllib.request import Request, urlopen
request = Request("https://superindex.app/api/v1/frontier.json",
headers={"User-Agent": "FrontierReader/1.0"})
with urlopen(request, timeout=30) as response:
data = json.load(response)
assert data["schema"] == "superindex.frontier/1"
for lab in data["labs"]:
print(lab["provider_name"], [m["id"] for m in lab["frontier"]]) Fetch https://superindex.app/api/v1/frontier.json; use ranked models for quality comparisons, preserve provisional/unknown values, check generated_at, and cite each model's sources plus SuperIndex. How each file is built
The examples below are generated from this build’s live snapshot. The frontier example shows the first lab; the models example shows one entry. The actual endpoints contain the complete collections.
frontier.json
Schema superindex.frontier/1. Each lab includes its top-ranked text model and other ranked models within 2 SI points of that leader, plus its newest unranked flagship candidate. Labs sort by their best rank; labs with no ranked models follow by provider slug.
“Flagship candidate” is a general name-based selection rule: it excludes explicit mini, small, nano, lite, micro, tiny, flash, Haiku, Sonnet, fast and lightning tiers. It is an inference for discovery, not a lab’s official designation. An unranked candidate has status: "provisional". latest_release is the newest dated text model of any tier. Missing or future dates do not establish the newest release; date ties use model ID. Month-only release dates retain their published precision, and exact dates in the same month sort ahead of them, since their true order within that month is unknown.
First lab in this build: Anthropic
{
"schema": "superindex.frontier/1",
"generated_at": "2026-10-09T06:15:14.937Z",
"source": "https://superindex.app",
"docs": "https://superindex.app/api/",
"license": "CC-BY-4.0",
"attribution": "SuperIndex — the Superintelligence Leaderboard",
"method_version": "si-v3-retained-evidence-2",
"methodology": "https://superindex.app/methodology/",
"price_unit": "USD per 1M tokens",
"selection": {
"si_margin": 2,
"flagship_rule": "Newest dated text model excluding explicit small/budget tiers; inferred from name, not an official lab designation",
"unknown_release_dates": "Excluded from newest-release selection",
"partial_release_dates": "Month-only dates retain source precision; exact dates in that month sort ahead of them, without asserting a precise release day"
},
"labs": [
{
"provider": "anthropic",
"provider_name": "Anthropic",
"frontier": [
{
"id": "anthropic/claude-fable-5-1",
"name": "Claude Fable 5.1",
"provider": "anthropic",
"provider_name": "Anthropic",
"release_date": "2026-09-01",
"release_date_precision": "day",
"modalities_in": [
"text",
"image",
"pdf"
],
"modalities_out": [
"text"
],
"open_weights": false,
"price_in": 10,
"price_out": 50,
"context_window": 1000000,
"si_score": 80.2,
"rank": 1,
"confidence_pct": 100,
"status": "ranked",
"provisional_reason": null,
"url": "https://superindex.app/models/anthropic/claude-fable-5-1/",
"sources": {
"price": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_in": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_out": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"release_date": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"context_window": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"benchmarks": [
{
"url": "https://arcprize.org/media/data/evaluations.json",
"retrieved_at": "2026-10-09T01:09:22.024Z"
},
{
"url": "https://epoch.ai/data/benchmark_data.zip",
"retrieved_at": "2026-10-09T01:09:16.471Z"
},
{
"url": "https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/resolve/main/text/latest-00000-of-00001.parquet",
"retrieved_at": "2026-10-09T04:32:42.300Z"
},
{
"url": "https://labs.scale.com/leaderboard/humanitys_last_exam",
"retrieved_at": "2026-10-09T01:09:22.845Z"
},
{
"url": "https://livebench.ai/table_2026_06_25.csv",
"retrieved_at": "2026-10-09T01:09:24.911Z"
},
{
"url": "https://www.anthropic.com/claude-fable-and-mythos-5-1",
"retrieved_at": "2026-10-09T01:09:14.679Z"
},
{
"url": "https://www.tbench.ai/leaderboard/terminal-bench/4.0",
"retrieved_at": "2026-10-09T01:29:27.569Z"
}
]
}
},
{
"id": "anthropic/claude-opus-5-5",
"name": "Claude Opus 5.5",
"provider": "anthropic",
"provider_name": "Anthropic",
"release_date": "2026-09-22",
"release_date_precision": "day",
"modalities_in": [
"text",
"image",
"pdf"
],
"modalities_out": [
"text"
],
"open_weights": false,
"price_in": 4,
"price_out": 20,
"context_window": 1000000,
"si_score": 78.4,
"rank": 2,
"confidence_pct": 100,
"status": "ranked",
"provisional_reason": null,
"url": "https://superindex.app/models/anthropic/claude-opus-5-5/",
"sources": {
"price": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_in": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_out": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"release_date": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"context_window": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"benchmarks": [
{
"url": "https://arcprize.org/media/data/evaluations.json",
"retrieved_at": "2026-10-09T01:09:22.024Z"
},
{
"url": "https://epoch.ai/data/benchmark_data.zip",
"retrieved_at": "2026-10-09T01:09:16.471Z"
},
{
"url": "https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/resolve/main/text/latest-00000-of-00001.parquet",
"retrieved_at": "2026-10-09T04:32:42.300Z"
},
{
"url": "https://livebench.ai/table_2026_06_25.csv",
"retrieved_at": "2026-10-09T01:09:24.911Z"
},
{
"url": "https://www.anthropic.com/claude-opus-5-5",
"retrieved_at": "2026-10-09T01:09:14.679Z"
},
{
"url": "https://www.tbench.ai/leaderboard/terminal-bench/4.0",
"retrieved_at": "2026-10-09T01:29:27.569Z"
}
]
}
}
],
"latest_release": {
"id": "anthropic/claude-haiku-5-5",
"name": "Claude Haiku 5.5",
"provider": "anthropic",
"provider_name": "Anthropic",
"release_date": "2026-10-07",
"release_date_precision": "day",
"modalities_in": [
"text",
"image",
"pdf"
],
"modalities_out": [
"text"
],
"open_weights": false,
"price_in": 0.1,
"price_out": 0.5,
"context_window": 1000000,
"si_score": 59.4,
"rank": 64,
"confidence_pct": 68,
"status": "ranked",
"provisional_reason": null,
"url": "https://superindex.app/models/anthropic/claude-haiku-5-5/",
"sources": {
"price": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_in": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_out": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"release_date": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"context_window": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"benchmarks": [
{
"url": "https://labs.scale.com/leaderboard/humanitys_last_exam",
"retrieved_at": "2026-10-09T01:09:22.845Z"
},
{
"url": "https://livebench.ai/table_2026_06_25.csv",
"retrieved_at": "2026-10-09T01:09:24.911Z"
},
{
"url": "https://www.anthropic.com/claude-haiku-5-5",
"retrieved_at": "2026-10-09T01:09:14.679Z"
}
]
}
}
}
]
}models.json
Schema superindex.models/1. Every text model: ranked first by rank, then provisional by newest release date, with unknown dates last. Image, video, speech, embedding and robotics products remain in the site catalog.
One model entry
{
"schema": "superindex.models/1",
"generated_at": "2026-10-09T06:15:14.937Z",
"source": "https://superindex.app",
"docs": "https://superindex.app/api/",
"license": "CC-BY-4.0",
"attribution": "SuperIndex — the Superintelligence Leaderboard",
"method_version": "si-v3-retained-evidence-2",
"methodology": "https://superindex.app/methodology/",
"price_unit": "USD per 1M tokens",
"models": [
{
"id": "anthropic/claude-fable-5-1",
"name": "Claude Fable 5.1",
"provider": "anthropic",
"provider_name": "Anthropic",
"release_date": "2026-09-01",
"release_date_precision": "day",
"modalities_in": [
"text",
"image",
"pdf"
],
"modalities_out": [
"text"
],
"open_weights": false,
"price_in": 10,
"price_out": 50,
"context_window": 1000000,
"si_score": 80.2,
"rank": 1,
"confidence_pct": 100,
"status": "ranked",
"provisional_reason": null,
"url": "https://superindex.app/models/anthropic/claude-fable-5-1/",
"sources": {
"price": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_in": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"price_out": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"release_date": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"context_window": {
"url": "https://platform.claude.com/docs/en/models/overview",
"retrieved_at": "2026-10-09T01:09:41.769Z"
},
"benchmarks": [
{
"url": "https://arcprize.org/media/data/evaluations.json",
"retrieved_at": "2026-10-09T01:09:22.024Z"
},
{
"url": "https://epoch.ai/data/benchmark_data.zip",
"retrieved_at": "2026-10-09T01:09:16.471Z"
},
{
"url": "https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/resolve/main/text/latest-00000-of-00001.parquet",
"retrieved_at": "2026-10-09T04:32:42.300Z"
},
{
"url": "https://labs.scale.com/leaderboard/humanitys_last_exam",
"retrieved_at": "2026-10-09T01:09:22.845Z"
},
{
"url": "https://livebench.ai/table_2026_06_25.csv",
"retrieved_at": "2026-10-09T01:09:24.911Z"
},
{
"url": "https://www.anthropic.com/claude-fable-and-mythos-5-1",
"retrieved_at": "2026-10-09T01:09:14.679Z"
},
{
"url": "https://www.tbench.ai/leaderboard/terminal-bench/4.0",
"retrieved_at": "2026-10-09T01:29:27.569Z"
}
]
}
}
]
}index.json
Schema superindex.index/1. Lists the data endpoints, schema IDs and JSON Schema URLs; docs links here.
The endpoint index
{
"schema": "superindex.index/1",
"generated_at": "2026-10-09T06:15:14.937Z",
"source": "https://superindex.app",
"docs": "https://superindex.app/api/",
"license": "CC-BY-4.0",
"attribution": "SuperIndex — the Superintelligence Leaderboard",
"method_version": "si-v3-retained-evidence-2",
"methodology": "https://superindex.app/methodology/",
"price_unit": "USD per 1M tokens",
"endpoints": [
{
"url": "https://superindex.app/api/v1/frontier.json",
"schema": "superindex.frontier/1",
"json_schema": "https://superindex.app/api/v1/frontier.schema.json"
},
{
"url": "https://superindex.app/api/v1/models.json",
"schema": "superindex.models/1",
"json_schema": "https://superindex.app/api/v1/models.schema.json"
}
]
}Fields and provenance
idnameproviderprovider_name- Stable canonical model ID and readable model/lab labels. Provider slugs are grouping keys, not display names.
release_daterelease_date_precision- Sourced release date (YYYY-MM, YYYY-MM-DD or ISO timestamp), or null when unknown. Precision is month, day, timestamp or null; a month-only source never becomes an invented exact day.
modalities_inmodalities_out- Input/output capability lists, or null when unknown. Only text-output leaderboard models enter this API. Multimodal input is allowed.
open_weights- True, false or null. Open weights do not by themselves imply an unrestricted license.
price_inprice_out- USD per one million tokens, using the catalog’s exact first-party/official price. Null means no verified rate; zero means a reported free rate. Never a sibling-model substitute.
context_window- Advertised input context in tokens, or null.
si_scorerankconfidence_pct- SI Score on a 0–100 scale, rank among eligible text models, and expected-evidence completeness on a 0–100 scale. Null rank means provisional even when a provisional score exists. Confidence is not answer accuracy.
statusprovisional_reason- ranked or provisional, with the reason when provisional. The API derives this status from rank eligibility.
url- Absolute SuperIndex model-page URL, or null when the model does not meet the page’s evidence rule.
sources.pricesources.release_date- Source URL and original retrieval time, or null. price uses input-price provenance first, otherwise output-price provenance.
sources.price_insources.price_outsources.context_window- Separate receipts for exact fields, particularly when input and output prices come from different pages or times.
sources.benchmarks- Deduplicated source URL/retrieval receipts for the model’s benchmark evidence. Detailed results remain on the linked model page.
generated_atmethod_versionmethodology- Snapshot generation time and score-method provenance. Individual receipts can be older: retained evidence is not made fresh by a refresh.
Updates, caching and license
The pipeline runs daily. A new snapshot publishes after the pipeline, site and live checks succeed. Sources publish at different times, so retained observations can be older than generated_at. This is a daily snapshot, not a real-time promise.
All version-1 files allow cross-origin browser reads with Access-Control-Allow-Origin: *. They use Cache-Control: public, max-age=0, must-revalidate, so caches revalidate. Conditional requests with the returned ETag can avoid downloading unchanged bytes. Refresh once per day or when your task needs a current snapshot.
Version 1 keeps the required field names and meanings stable. Additive fields may appear; ignore fields your client does not use. Breaking changes get a new versioned path and schema ID. JSON Schemas use Draft 2020-12.
SuperIndex’s dataset and composite output are CC BY 4.0. Attribute SuperIndex — the Superintelligence Leaderboard, link to superindex.app, retain source receipts and indicate your modifications. Individual benchmark publishers retain their own source licenses; the cited facts do not transfer ownership of their pages.
For explanations and detailed benchmark evidence, use the methodology and each model’s sources. The Atom feed and change history provide a separate record of leaderboard changes.