GPT-Rosalind

Fiche modèle · OpenAI · gpt-rosalind-research

GPT-Rosalind is OpenAI's specialized life-sciences reasoning model for approved internal research. OpenAI introduced GPT-Rosalind in research preview on April 16, 2026 and updated the same launch materials on September 11, 2026 to mark general availability through the trusted-access program for eligible Enterprise and Business organizations. Official OpenAI docs describe it as the company's most capable model for life sciences research, combining deep scientific reasoning with scientific tool and database use for multi-step workflows across target research, genomics and sequencing analysis, protein and sequence analysis, medicinal chemistry, literature synthesis, wet-lab troubleshooting, and experiment planning. OpenAI publicly lists API pricing for the `gpt-rosalind-research` model id at $5 input, $0.50 cached input, and $25 output per million tokens, with no cache-write pricing. Benchmarks below are taken from OpenAI's official GPT-Rosalind-5.5 system card and LifeSciBench preprint, the latest first-party benchmark sources published for the GPT-Rosalind series. OpenAI does not publish a dedicated model-reference page with context-window or maximum-output limits for GPT-Rosalind, so those fields remain null.

Identité

OrganisationOpenAI
Pays—
Famille—
Typellm
Date de sortie2026-09-11
Licenceproprietary
Licence commercialenon
Licence open sourcenon
Poids ouvertsnon
Auto-hébergement (DGX Spark)non — poids fermés
Modalitésimage|text
Fenêtre de contexte—
KV cache / token (Q8)—
Paramètres totaux—
Paramètres actifs (session)—
Source des paramètres actifs—
Curation des paramètres actifs—
Prix par session10.00 $ (1 000 000 tokens entrée + 200 000 sortie, prix le moins cher par fournisseur)
Tokens / tâche (Data Players)1.2 M
Tokens / tâche (Terminal-Bench)83.3 M (est.)
Ratio Terminal-Bench / Data Players69.4×
Coût / tâche (Terminal-Bench)83.98 $ (est.)
Tokens d'entraînement—
Coupure de connaissance—
Créé le2026-10-07T21:38:19.234531Z
Mis à jour le2026-11-04T06:05:03.726147Z
SourceLLM Stats
URLhttps://llm-stats.com/models/gpt-rosalind-research

Prix par fournisseur

Statut du modèle (offres + auto-hébergement) : non.

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 « — ».

FournisseurSourceSouverainetéEntrée (/M tok)Sortie (/M tok)QuantificationContexteLatence (s)Débit (tok/s)Statut
OpenAIllm-statsnon5.00 $25.00 $————active

Souveraineté qualifiée par offre (modèle × fournisseur) ; survolez le badge pour les critères.

Souveraineté

Scores par catégorie (top_scores)

CatégorieScore
multimodal0.60
reasoning0.40
science0.40
general0.40
healthcare0.40
agents0.40
vision0.60

Scores par benchmark

BenchmarkCatégorieScoreMaxNormalisé (API)VérifiéAuto-déclaréDaté du
biorisk-knowledgehealthcare0.811.000.81nonoui2026-11-04T06:05:03.767081Z
dna-sequence-design-for-transcription-factor-bindinghealthcare0.141.000.14nonoui2026-11-04T06:05:03.787145Z
genebenchagents0.221.000.22nonoui2026-11-04T06:05:03.737604Z
hard-negative-protein-binding-predictionagents0.031.000.03nonoui2026-11-04T06:05:03.784010Z
labworkbenchagents0.631.000.63nonoui2026-11-04T06:05:03.745180Z
lifescibenchagents0.581.000.58nonoui2026-11-04T06:05:03.741624Z
medchembenchreasoning0.281.000.28nonoui2026-11-04T06:05:03.732706Z
multi-select-virology-troubleshootinghealthcare0.511.000.51nonoui2026-11-04T06:05:03.780811Z
protocolqa-open-endedsafety0.361.000.36nonoui2026-11-04T06:05:03.749626Z
troubleshootingbenchhealthcare0.531.000.53nonoui2026-11-04T06:05:03.757743Z

Classements

CatégorieRangScoreNote conservatriceBenchmarks évaluésMéthodeDaté duProvenance
agents72—21.542trueskill2026-10-19model_history

Lecture : Note conservatrice = borne basse TrueSkill (niveau estimé − marge d’incertitude) ; Benchmarks évalués = nombre de benchmarks de la catégorie pris en compte. À performances égales, un modèle évalué sur moins de benchmarks obtient une note plus basse : la note ne se déduit pas d’un seul score. Provenance : rankings = top 50 de /rankings ; model_history = historique du modèle (/models/{id}/history), utilisé pour les rangs au-delà du top 50.

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