Claude Mythos Preview
Claude Mythos Preview is an unreleased general-purpose frontier model from Anthropic, a new tier above Opus (internal codename 'Capybara'). It identified thousands of zero-day vulnerabilities across every major operating system and web browser as part of Project Glasswing, a cross-industry cybersecurity initiative with 12 partners including AWS, Apple, Microsoft, and Google. State-of-the-art on SWE-bench Verified (93.9%), GPQA Diamond (94.6%), USAMO (97.6%), Terminal-Bench 2.0 (82.0%), CyberGym (83.1%), and Cybench (100% pass@1, saturated). Represents a 4.3x increase over the previous trendline for model performance. Deployed under ASL-3 Standard. Best-aligned Claude model to date per Anthropic's risk report, with the first-ever 24-hour internal alignment review before deployment. Not planned for general availability. Pricing for participants: $25/$125 per million tokens (input/output). 244-page system card.
Identité
| Organisation | Anthropic |
|---|---|
| Pays | — |
| Famille | — |
| Type | llm |
| Date de sortie | — |
| Licence | proprietary |
| Licence commerciale | non |
| Licence open source | non |
| Poids ouverts | non |
| Auto-hébergement (DGX Spark) | non — poids fermés |
| Modalités | image|text |
| Fenêtre de contexte | 1 000 000 |
| KV cache / token (Q8) | — |
| Paramètres totaux | — |
| Paramètres actifs (session) | — |
| Source des paramètres actifs | — |
| Curation des paramètres actifs | — |
| Prix par session | — |
| Tokens / tâche (Data Players) | 1.2 M |
| Tokens / tâche (Terminal-Bench) | 83.3 M (est.) |
| Ratio Terminal-Bench / Data Players | 69.4× |
| Coût / tâche (Terminal-Bench) | — |
| Tokens d'entraînement | — |
| Coupure de connaissance | — |
| Créé le | 2026-05-07T16:53:27.671423Z |
| Mis à jour le | 2026-11-04T06:04:32.980137Z |
| Source | LLM Stats |
| URL | https://llm-stats.com/models/claude-mythos-preview |
Prix par fournisseur
Toutes sources confondues : catalogue llm-stats, routeurs EUrouter/Cortecs et catalogues directs (OVHcloud, Scaleway, Infomaniak). Les prix sont convertis en USD pour la comparaison ; le prix publié et le taux BCE utilisé sont dans l'infobulle. Les prix inconnus restent indiqués « — ».
Aucune donnée.
Souveraineté qualifiée par offre (modèle × fournisseur) ; survolez le badge pour les critères.
SouverainetéScores par catégorie (top_scores)
| Catégorie | Score |
|---|---|
| language | 0.90 |
| long_context | 0.80 |
| math | 0.80 |
| multimodal | 0.80 |
| physics | 0.90 |
| reasoning | 0.80 |
| safety | 0.90 |
| search | 0.90 |
| spatial_reasoning | 0.80 |
| frontend_development | 0.90 |
| general | 0.90 |
| healthcare | 0.90 |
| agents | 0.80 |
| biology | 0.90 |
| chemistry | 0.90 |
| code | 0.80 |
| tool_calling | 0.80 |
| vision | 0.80 |
Scores par benchmark
| Benchmark | Catégorie | Score | Max | Normalisé (API) | Vérifié | Auto-déclaré | Daté du |
|---|---|---|---|---|---|---|---|
| browsecomp | reasoning | 0.87 | 1.00 | 0.87 | non | oui | 2026-11-04T06:04:33.053175Z |
| charxiv-r | reasoning | 0.93 | 1.00 | 0.93 | non | oui | 2026-11-04T06:04:33.093146Z |
| cybench | safety | 1.00 | 1.00 | 1.00 | non | oui | 2026-11-04T06:04:33.104599Z |
| cybergym | code | 0.83 | 1.00 | 0.83 | non | oui | 2026-11-04T06:04:33.063323Z |
| figqa | safety | 0.89 | 1.00 | 0.89 | non | oui | 2026-11-04T06:04:33.100542Z |
| gpqa | reasoning | 0.95 | 1.00 | 0.95 | non | oui | 2026-11-04T06:04:33.030180Z |
| graphwalks-bfs->128k | reasoning | 0.80 | 1.00 | 0.80 | non | oui | 2026-11-04T06:04:33.080976Z |
| humanity's-last-exam | reasoning | 0.65 | 1.00 | 0.65 | non | oui | 2026-11-04T06:04:33.039981Z |
| mmmlu | language | 0.93 | 1.00 | 0.93 | non | oui | 2026-11-04T06:04:33.074582Z |
| osworld-verified | multimodal | 0.80 | 1.00 | 0.80 | non | oui | 2026-11-04T06:04:33.059110Z |
| swe-bench-multilingual | reasoning | 0.87 | 1.00 | 0.87 | non | oui | 2026-11-04T06:04:33.008846Z |
| swe-bench-multimodal | reasoning | 0.59 | 1.00 | 0.59 | non | oui | 2026-11-04T06:04:33.013215Z |
| swe-bench-pro | reasoning | 0.78 | 1.00 | 0.78 | non | oui | 2026-11-04T06:04:33.003866Z |
| swe-bench-verified | reasoning | 0.94 | 1.00 | 0.94 | non | oui | 2026-11-04T06:04:32.991731Z |
| terminal-bench-2 | reasoning | 0.82 | 1.00 | 0.82 | non | oui | 2026-11-04T06:04:33.023355Z |
| usamo25 | math | 0.98 | 1.00 | 0.98 | non | oui | 2026-11-04T06:04:33.069223Z |
Classements
| Catégorie | Rang | Score | Note conservatrice | Benchmarks évalués | Méthode | Daté du | Provenance |
|---|---|---|---|---|---|---|---|
| agents | 22 | 0.80 | 36.51 | 7 | trueskill | 2026-11-04T08:08:47.839150Z | rankings |
| code | 5 | 0.80 | 44.20 | 7 | trueskill | 2026-11-04T08:08:39.742311Z | rankings |
| math | 1 | 0.80 | 47.63 | 3 | trueskill | 2026-11-04T08:08:42.442591Z | rankings |
| multimodal | 7 | 0.80 | 36.94 | 3 | trueskill | 2026-11-04T08:08:46.488887Z | rankings |
| reasoning | 4 | 0.80 | 55.11 | 12 | trueskill | 2026-11-04T08:08:41.093494Z | rankings |
Données
Dates de curation (récupération depuis l'API) : models 2026-11-04 · benchmarks 2026-11-04 · scores 2026-11-04 · updates 2026-11-04 · account 2026-11-04 · model_history 2026-11-04 · rankings 2026-11-04