Mistral
Île-de-France, FrançaSobre Mistral
Mistral is an AI model company with registered offices in Paris, France. Their mission is to democratise artificial intelligence by building efficient, open-source and high-performance models which organisations can deploy and own.
Desempenho
Desempenho recente do Gateway nos modelos de Mistral.
—
Sem dados recentes
—
Sem dados recentes
—
Sem dados recentes
As métricas de desempenho aparecerão quando os modelos deste laboratório processarem tráfego do Gateway.
Modelos recentes
Lançamentos recentes com detalhes de disponibilidade, contexto, preços e uso.
Mistral Large 4 is Mistral AI's open-weight flagship: a mixture-of-experts model with about 1.05T total and 49B active parameters, a 1.6B vision encoder, image and text input, a 1M-token context and low, medium or high reasoning. It is the improved release that replaced the October 2026 preview.
Voxtral Mini Transcribe 3 is Mistral's speech-to-text model for transcription with speaker diarization and timestamps at a low per-minute price. It is suited to captioning, meeting notes, call analytics, and batch transcription pipelines. It is in public preview.
Voxtral Mini Transcribe Realtime 3 is Mistral's streaming speech-to-text model for agentic and live-audio use. It has adjustable latency from 80 ms to 2,400 ms, built-in end-of-turn detection, context biasing, and speaker diarization in a single WebSocket session. It is in public preview.
Voxtral Small Transcribe 3 is Mistral's flagship speech-to-text model, with speaker diarization and timestamps. It also handles translation, speech Q&A, summarization, and function calling. It is in public preview.
Voxtral TTS 3 is Mistral's text-to-speech model with zero-shot and professional voice cloning across 9 languages. It streams audio with a time to first audio of about 90 ms and supports 34 emotion tags and 2 pace tags. It is in public preview.
Mistral Large 4 Preview was the public preview of Mistral Large 4, available on the Mistral API from 6 October 2026. On 29 October 2026 Mistral replaced it in place with the improved open-weight Mistral Large 4.
OCR 4.1 is Mistral's document understanding model with improved layout fidelity, block detection, text preservation, checkbox recognition, and right-to-left table parsing.
Shieldstral 1.0 3B is Mistral's open-weight, policy-adaptive multimodal safety classifier. It evaluates text, images, or combined text-and-image content against natural-language safety policies and returns a single yes/no verdict with a confidence score.
Leanstral 1.5 is a Mistral Labs model optimized for Lean 4 formal proof engineering, automated theorem proving, and autoformalization. Mistral documents labs-leanstral-1-5 as the public API model slug.
OCR 4 is Mistral's current document understanding model for OCR, extraction, and page-level visual reasoning across PDFs, scans, and screenshots.
Mistral Medium 3.5 is a multimodal model from Mistral that can work across text and visual inputs. It accepts text and image inputs and produces text outputs.
Voxtral TTS is Mistral's text-to-speech model with zero-shot voice cloning. It accepts text plus a short reference audio prompt and produces natural-sounding speech output.
Na web
Perfis oficiais, pesquisas e canais da comunidade.
