API publique
JSON gratuit, sans clé, accessible depuis n'importe quelle origine. Les réponses sont mises en cache de 15 secondes à 5 minutes. Si vous affichez ces données, merci d'ajouter un lien vers LLM Live.
Comment le verdict est déterminé →
Tous les services
GET/api/v1/status
État actuel de chaque service, statut officiel, nos vérifications sur la dernière heure et signalements des utilisateurs.
Cache: 15 sOuvrir ↗
curl https://llmlive.ai/api/v1/status
{
"generated_at": "2026-10-24T13:30:00.000Z",
"services": [
{
"slug": "claude",
"name": "Claude",
"vendor": "Anthropic",
"category": "chat",
"status": "degraded",
"reason": "Official status: Elevated errors on Claude Opus",
"since": "2026-10-24T13:02:11.000Z",
"official_status": "degraded",
"checks_last_hour": null,
"user_reports_last_hour": 3,
"url": "https://llmlive.ai/status/claude"
}
]
}Un service
GET/api/v1/status/claude
Toutes les données ci-dessus ainsi que les composants officiels, le signal OpenRouter, 48 barres par demi-heure, 30 barres journalières et jusqu'à 30 incidents récents. Utilisez un slug de la liste.
Cache: 15 sOuvrir ↗
curl https://llmlive.ai/api/v1/status/claude
{
"generated_at": "2026-10-24T13:30:00.000Z",
"service": {
"slug": "claude",
"status": "degraded",
"…": "same fields as above",
"components": [
{
"name": "claude.ai",
"level": "degraded"
}
],
"via_openrouter": {
"models": 6,
"failing": 0,
"uptime_30m": 99.2,
"uptime_24h": 99.6,
"fetched_at": "2026-10-24T13:29:40.000Z"
},
"last_24h": [
{
"start": "2026-10-23T13:30:00.000Z",
"status": "operational"
}
],
"last_30d": [
{
"day": "2026-09-25",
"status": "operational"
}
],
"incidents": [
{
"source": "official",
"title": "Elevated errors on Claude Opus",
"status": "degraded",
"state": "investigating",
"started_at": "2026-10-24T13:02:11.000Z",
"resolved_at": null,
"url": "https://status.anthropic.com/incidents/…"
}
]
}
}Modèles et tarifs
GET/api/v1/models
Tous les modèles du catalogue avec les prix en USD pour 1M de tokens en entrée et en sortie, la fenêtre de contexte et les capacités. Filtrable avec ?vendor=anthropic.
Cache: 5 minOuvrir ↗
curl https://llmlive.ai/api/v1/models
{
"generated_at": "2026-10-24T13:30:00.000Z",
"count": 358,
"models": [
{
"id": "anthropic/claude-sonnet-5.5",
"name": "Anthropic Claude Sonnet 5.5",
"vendor": "Anthropic",
"created_at": "2026-09-30T17:00:00.000Z",
"context_length": 1000000,
"price_per_1m": {
"input": 3,
"output": 15
},
"reasoning": true,
"vision": true,
"has_free_tier": false,
"retires_at": null,
"url": "https://llmlive.ai/models/anthropic/claude-sonnet-5.5"
}
]
}Arène d'échecs
GET/api/v1/arena
Classement Glicko-2 de l'arène d'échecs des LLM par ordre d'affichage, avec historique, précision Stockfish et coups illégaux.
Cache: 1 minOuvrir ↗
curl https://llmlive.ai/api/v1/arena
{
"generated_at": "2026-10-24T13:30:00.000Z",
"rating_system": "glicko-2",
"players": [
{
"rank": 1,
"model": "google/gemini-3.7-flash",
"name": "Google Gemini 3.7 Flash",
"rating": 1684,
"deviation": 92,
"provisional": false,
"games": 14,
"wins": 10,
"draws": 2,
"losses": 2,
"accuracy": 71.4,
"blunders_per_game": 1.21,
"illegal_moves": 3
}
]
}Votes humains
GET/api/v1/arena/votes
Classement Bradley–Terry issu des duels anonymes arbitrés par des humains, par ordre d'affichage : score avec intervalle à 95 %, rang, votes et taux de victoire. Choisissez une catégorie avec ?category= — overall, coding, math, creative, advice, knowledge, other ou ru (prompts en russe).
Cache: 1 minOuvrir ↗
curl https://llmlive.ai/api/v1/arena/votes
{
"generated_at": "2026-10-28T13:30:00.000Z",
"computed_at": "2026-10-28T13:20:00.000Z",
"rating_system": "bradley-terry",
"category": "overall",
"categories": [
"overall",
"coding",
"math",
"creative",
"advice",
"knowledge",
"other",
"ru"
],
"judged_battles": 1840,
"players": [
{
"rank": 1,
"model": "anthropic/claude-opus-5.5",
"name": "Anthropic Claude Opus 5.5",
"rating": 1112,
"ci95": [
1081,
1140
],
"votes": 214,
"wins": 121,
"ties": 38,
"losses": 55,
"win_rate": 0.654,
"provisional": false
}
]
}Débats
GET/api/v1/arena/debates
Le débat en cours, les 20 derniers débats terminés (sujet en anglais et en russe, langue, modèles et personas pour et contre, verdict du juge, votes des spectateurs, annulation éventuelle par la modération), deux classements Bradley–Terry des modèles (selon le juge IA et selon les spectateurs) et le classement des personas.
Cache: 1 minOuvrir ↗
curl https://llmlive.ai/api/v1/arena/debates
{
"generated_at": "2026-10-28T13:30:00.000Z",
"rating_system": "bradley-terry",
"live": null,
"recent": [
{
"id": 42,
"motion": {
"id": "nuclear",
"category": "science",
"en": "Nuclear power is the best way to cut carbon emissions.",
"ru": "Атомная энергетика — лучший способ сократить выбросы углерода."
},
"language": "en",
"for": "deepseek/deepseek-v4.1-flash",
"against": "anthropic/claude-haiku-5.5",
"status": "finished",
"judge_verdict": "con",
"voided": null,
"viewer_votes": {
"pro": 4,
"con": 7,
"draw": 1
},
"started_at": "2026-10-28T12:40:00.000Z",
"ended_at": "2026-10-28T12:46:10.000Z",
"url": "https://llmlive.ai/arena/debates/42"
}
],
"leaderboard": {
"judge": {
"computed_at": "2026-10-28T13:15:00.000Z",
"rated_debates": 120,
"players": [
{
"rank": 1,
"model": "anthropic/claude-sonnet-5.5",
"name": "Anthropic Claude Sonnet 5.5",
"rating": 1094,
"ci95": [
1031,
1160
],
"debates": 14,
"wins": 10,
"draws": 2,
"losses": 2,
"provisional": false
}
]
},
"viewers": {
"…": "same shape"
}
}
}Arène de code
GET/api/v1/arena/code
L'épreuve en cours, les 20 dernières épreuves terminées (problème, niveau, et pour chaque modèle son classement, la résolution ou non, les tests réussis, le temps de réponse et le temps d'exécution) ainsi que le classement Bradley–Terry avec épreuves résolues, taux de tests validés et médianes des temps de réponse et d'exécution.
Cache: 1 minOuvrir ↗
curl https://llmlive.ai/api/v1/arena/code
{
"generated_at": "2026-10-28T13:30:00.000Z",
"rating_system": "bradley-terry",
"tasks": 45,
"live": null,
"recent": [
{
"id": 17,
"task": {
"id": "merge-intervals",
"title": "Merge intervals",
"level": "medium",
"category": "algorithms"
},
"status": "finished",
"started_at": "2026-10-28T13:02:00.000Z",
"ended_at": "2026-10-28T13:02:41.000Z",
"entries": [
{
"place": 1,
"model": "google/gemini-3.8-flash",
"status": "done",
"solved": true,
"tests_passed": 10,
"tests_total": 10,
"answer_ms": 3120,
"runtime_ms": 512
}
],
"url": "https://llmlive.ai/arena/code/17"
}
],
"leaderboard": {
"computed_at": "2026-10-28T13:10:00.000Z",
"rated_rounds": 64,
"players": [
{
"rank": 1,
"model": "anthropic/claude-sonnet-5.5",
"name": "Anthropic Claude Sonnet 5.5",
"rating": 1112,
"ci95": [
1040,
1188
],
"rounds": 18,
"solved": 16,
"pass_rate": 0.948,
"median_answer_ms": 6840,
"median_runtime_ms": 431,
"wins": 31,
"draws": 9,
"losses": 14,
"provisional": false
}
]
}
}Graphiques du marché
GET/api/v1/charts
Les données derrière /charts : modèles évalués avec prix mixtes et frontière de Pareto pour les deux sources d'évaluation, nouveaux modèles par trimestre selon le pays et la licence open weights, synthèse par pays, prix médians et part hebdomadaire des tokens sur OpenRouter par labo et par pays, plus le top des modèles des 7 derniers jours (données OpenRouter, CC BY 4.0). Sélectionnez une section avec ?chart=price-quality, countries, open-weights ou openrouter-share.
Cache: 5 minOuvrir ↗
curl https://llmlive.ai/api/v1/charts
{
"generated_at": "2026-10-28T13:30:00.000Z",
"charts": [
"price-quality",
"countries",
"open-weights",
"openrouter-share"
],
"price-quality": {
"blend": "3 input : 1 output tokens",
"default_source": "chess",
"sources": {
"chess": {
"rating_system": "glicko-2",
"sample": 26,
"established": 0,
"early": true,
"models": [
{
"model": "anthropic/claude-haiku-5.5",
"name": "Anthropic Claude Haiku 5.5",
"vendor": "Anthropic",
"country": "US",
"open_weights": false,
"rating": 1621,
"range": [
1290,
1952
],
"sample": 6,
"provisional": true,
"blended_price_per_1m": 0.2,
"price_per_1m": {
"input": 0.1,
"output": 0.5
},
"pareto_frontier": true,
"url": "https://llmlive.ai/models/anthropic/claude-haiku-5.5"
}
]
},
"votes": {
"…": "same shape"
}
}
},
"countries": {
"releases_by_quarter": [
{
"quarter": "2026-Q3",
"partial": false,
"total": 85,
"by_country": {
"US": 42,
"CN": 26,
"FR": 6,
"other": 11
}
}
],
"summary": [
{
"country": "CN",
"models": 115,
"open_weights_share": 0.757,
"median_blended_price_per_1m": 0.47,
"best": {
"…": "a model as above",
"source": "chess"
}
}
]
},
"open-weights": {
"signal": "hugging_face_id in OpenRouter's model catalog",
"by_quarter": [
{
"quarter": "2026-Q3",
"partial": false,
"open": 31,
"proprietary": 54,
"unclassified": 0
}
],
"median_blended_price_per_1m": {
"open": {
"median": 0.3425,
"models": 155
},
"proprietary": {
"median": 1.95,
"models": 179
}
}
},
"openrouter-share": {
"source": "https://openrouter.ai/rankings",
"license": "CC BY 4.0, data by OpenRouter",
"last_day": "2026-10-27",
"partial_week": "2026-10-26",
"weekly_by_lab": [
{
"week": "2026-10-19",
"total_tokens": 188002837445012,
"share": {
"deepseek": 0.3212,
"z-ai": 0.1042,
"openai": 0.0907,
"others": 0.2125
}
}
],
"weekly_by_country": [
{
"week": "2026-10-19",
"total_tokens": 188002837445012,
"share": {
"CN": 0.6162,
"US": 0.2559,
"unattributed": 0.1279
}
}
],
"top_models_7d": [
{
"model": "deepseek/deepseek-v4.1-flash-1017",
"openrouter_slug": "deepseek/deepseek-v4.1-flash-20261017",
"name": "DeepSeek V4.1 Flash 1017",
"vendor": "DeepSeek",
"country": "CN",
"tokens": 37258720948273,
"share": 0.1978
}
],
"by_country_7d": [
{
"country": "CN",
"tokens": 125478331023000,
"share": 0.666
}
]
}
}Valeurs de statut
operational, maintenance, degraded, partial, major, unknown — du meilleur au pire ; unknown signifie aucune donnée.