MAI-Thinking-1
MAI-Thinking-1 is Microsoft AI's first in-house reasoning model, a 35B-active / ~1T-total parameter sparse Mixture of Experts model (base model MAI-Base-1) trained from scratch without distillation from third-party models. Built with Microsoft's Hill-Climbing Machine pipeline, it was pre-trained on 30T tokens of clean, commercially licensed, human-generated data (plus 3.55T mid-training tokens), then post-trained via reinforcement learning across STEM, agentic coding, and helpfulness/safety specialists consolidated into a single model. It delivers strong mathematical reasoning and software-engineering performance for its weight class, going toe-to-toe with Claude Opus 4.6 on SWE-Bench Pro and reaching 97.0% on AIME 2025. It supports a 256k token context window, function calling, and developer instructions, and is preferred over Claude Sonnet 4.6 in blind human side-by-side evaluations.
Identité
| Organisation | Microsoft |
|---|---|
| Pays | — |
| Famille | — |
| Type | llm |
| Date de sortie | 2026-06-02 |
| Licence | proprietary |
| Licence commerciale | non |
| Licence open source | non |
| Poids ouverts | non |
| Auto-hébergement (DGX Spark) | non — poids fermés |
| Modalités | — |
| Fenêtre de contexte | 256 000 |
| KV cache / token (Q8) | — |
| Paramètres totaux | 1 T |
| Paramètres actifs (session) | 35 B |
| Source des paramètres actifs | web : https://microsoft.ai/pdf/mai-thinking-1.pdf |
| Curation des paramètres actifs | 2026-10-19 |
| 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 | 30 000 000 000 000 |
| Coupure de connaissance | — |
| Créé le | 2026-06-05T21:54:38.998093Z |
| Mis à jour le | 2026-11-04T06:04:52.903647Z |
| Source | LLM Stats |
| URL | https://llm-stats.com/models/mai-thinking-1 |
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 |
|---|---|
| chat | 0.50 |
| instruction_following | 0.70 |
| language | 0.80 |
| legal | 0.90 |
| long_context | 0.80 |
| math | 0.90 |
| multimodal | 0.40 |
| physics | 0.80 |
| reasoning | 0.70 |
| structured_output | 0.70 |
| factuality | 0.30 |
| finance | 0.80 |
| frontend_development | 0.70 |
| general | 0.70 |
| healthcare | 0.60 |
| agents | 0.60 |
| biology | 0.80 |
| chemistry | 0.80 |
| code | 0.60 |
| communication | 0.50 |
| tool_calling | 0.60 |
| vision | 0.40 |
Scores par benchmark
| Benchmark | Catégorie | Score | Max | Normalisé (API) | Vérifié | Auto-déclaré | Daté du |
|---|---|---|---|---|---|---|---|
| advancedif | instruction_following | 0.85 | 1.00 | 0.85 | non | oui | 2026-11-04T06:04:53.023217Z |
| aime-2025 | math | 0.97 | 1.00 | 0.97 | non | oui | 2026-11-04T06:04:52.913398Z |
| aime-2026 | math | 0.94 | 1.00 | 0.94 | non | oui | 2026-11-04T06:04:52.924685Z |
| air-bench | safety | 0.88 | 1.00 | 0.88 | non | oui | 2026-11-04T06:04:53.061033Z |
| bfcl-v3 | general | 0.72 | 1.00 | 0.72 | non | oui | 2026-11-04T06:04:53.053065Z |
| corpusqa | long_context | 0.82 | 1.00 | 0.82 | non | oui | 2026-11-04T06:04:53.143155Z |
| cyberseceval-4 | safety | 0.63 | 1.00 | 0.63 | non | oui | 2026-11-04T06:04:53.082400Z |
| gpqa | reasoning | 0.84 | 1.00 | 0.84 | non | oui | 2026-11-04T06:04:52.944416Z |
| graphwalks | long_context | 0.90 | 1.00 | 0.90 | non | oui | 2026-11-04T06:04:53.043186Z |
| healthbench-professional | healthcare | 0.35 | 1.00 | 0.35 | non | oui | 2026-11-04T06:04:53.113166Z |
| hmmt-feb-26 | math | 0.85 | 1.00 | 0.85 | non | oui | 2026-11-04T06:04:52.933220Z |
| ifbench | instruction_following | 0.69 | 1.00 | 0.69 | non | oui | 2026-11-04T06:04:53.013203Z |
| livecodebench-v6 | reasoning | 0.88 | 1.00 | 0.88 | non | oui | 2026-11-04T06:04:52.953474Z |
| longbench-v2 | long_context | 0.61 | 1.00 | 0.61 | non | oui | 2026-11-04T06:04:53.133133Z |
| longfact | factuality | 0.98 | 1.00 | 0.98 | non | oui | 2026-11-04T06:04:53.092473Z |
| medxpertqa | healthcare | 0.43 | 1.00 | 0.43 | non | oui | 2026-11-04T06:04:53.123542Z |
| mmlu-pro | language | 0.85 | 1.00 | 0.85 | non | oui | 2026-11-04T06:04:52.993338Z |
| multichallenge | chat | 0.53 | 1.00 | 0.53 | non | oui | 2026-11-04T06:04:53.033384Z |
| simpleqa-verified | factuality | 0.31 | 1.00 | 0.31 | non | oui | 2026-11-04T06:04:53.003180Z |
| swe-bench-pro | reasoning | 0.53 | 1.00 | 0.53 | non | oui | 2026-11-04T06:04:52.981252Z |
| swe-bench-verified | reasoning | 0.73 | 1.00 | 0.73 | non | oui | 2026-11-04T06:04:52.973359Z |
| terminal-bench-2 | reasoning | 0.46 | 1.00 | 0.46 | non | oui | 2026-11-04T06:04:52.966608Z |
| truthfulqa | general | 0.88 | 1.00 | 0.88 | non | oui | 2026-11-04T06:04:53.103527Z |
Classements
| Catégorie | Rang | Score | Note conservatrice | Benchmarks évalués | Méthode | Daté du | Provenance |
|---|---|---|---|---|---|---|---|
| agents | 113 | — | 12.60 | 3 | trueskill | 2026-10-19 | model_history |
| code | 120 | — | 19.05 | 3 | trueskill | 2026-10-19 | model_history |
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