Title: Phi-4-multimodal-instruct | LLM EU Canonical: https://llmeu.com/pt/models/phi-4-multimodal-instruct/ Markdown mirror: https://llmeu.com/pt/models/phi-4-multimodal-instruct.md Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json Provider: Microsoft Licence: mit Hosting: catalogue only %E2%80%94 not hosted on LLM EU yet Sources: https://huggingface.co/microsoft/Phi-4-multimodal-instruct https://arxiv.org/abs/2503.01743 # Phi-4-multimodal-instruct [Microsoft](/pt/providers/microsoft/) %C3%9Altima revis%C3%A3o: 14/10/2026 ## Soberania e alojamento Este modelo est%C3%A1 catalogado para refer%C3%AAncia. Ainda n%C3%A3o est%C3%A1 alojado na LLM EU, e nenhum endpoint pr%C3%B3prio o serve aqui. Os pedidos s%C3%A3o encaminhados diretamente para o fornecedor, sob os termos desse fornecedor. ## Resumo Phi-4-multimodal-instruct %C3%A9 um modelo de 5.6B que aceita texto, imagens e fala em um %C3%BAnico checkpoint, com um contexto de 128K tokens e a licen%C3%A7a MIT. O cart%C3%A3o do modelo reporta 55.1 no MMMU, 93.2 no DocVQA e 62.4 no MathVista, e lista os idiomas suportados separadamente para cada modalidade de entrada. %C3%89 uma op%C3%A7%C3%A3o compacta para pipelines de documentos e fala que, de outra forma, precisariam de dois modelos. ## Especifica%C3%A7%C3%B5es - Fornecedor Microsoft (US) - Janela de contexto 131k - Par%C3%A2metros 5.57B - Modalidades text, vision, audio - Licen%C3%A7a mit %C2%B7 [licence](https://opensource.org/license/mit) - Pesos abertos yes - Data de lan%C3%A7amento 24/02/2025 - Tarefas [Vis%C3%A3o](/pt/tasks/vision/), [Sumariza%C3%A7%C3%A3o](/pt/tasks/summarization/), [Suporte multilingue](/pt/tasks/multilingual-support/), [Baixo custo](/pt/tasks/cheap/) ## Idiomas **Forte:** en, de, fr, es, it, pt **Adequado:** nl, pl, sv, da, fi, cs, ru, uk, ar, zh, ja, ko, hi, tr #### Forte - en - de - fr - es - it - pt #### Adequado - nl - pl - sv - da - fi - cs - ru - uk - ar - zh - ja - ko - hi - tr N%C3%A3o verificado: o fornecedor n%C3%A3o documentou esta afirma%C3%A7%C3%A3o. %E2%80%94 a language is listed as strong only where a source claims it; the card links the evidence where one exists. ## Quando usar #### Pontos fortes - text, image and speech input at 5.6B parameters - 128K-token context for document batches - MIT licence - 23 declared languages including most large EU languages #### Limita%C3%A7%C3%B5es - modality-specific quality is below dedicated single-modality models - English-centric training data - no evaluation has been run by LLM EU ## Perguntas frequentes ### Is Phi-4-multimodal-instruct served on LLM EU? It has no endpoint in the catalogue, so it is catalogue only %E2%80%94 not hosted on LLM EU yet, with no region and no retention to report. A card whose status is "listed" is a catalogue entry, not a running deployment. ### Under what licence is Phi-4-multimodal-instruct released? Phi-4-multimodal-instruct is recorded under mit. The recorded licence text is https://opensource.org/license/mit. The weights are downloadable (open_weights is true). LLM EU republishes the licence id and the link and gives no legal advice; the licence text governs. ### How large is Phi-4-multimodal-instruct, and what context window does it have? Phi-4-multimodal-instruct is recorded at 5.57B (safetensors total 5,574,460,384) (source: https://huggingface.co/api/models/microsoft/Phi-4-multimodal-instruct), with a context window of 131072 tokens (source: https://huggingface.co/microsoft/Phi-4-multimodal-instruct). ### Pre%C3%A7os Este modelo est%C3%A1 catalogado para refer%C3%AAncia. Ainda n%C3%A3o est%C3%A1 alojado na LLM EU, e nenhum endpoint pr%C3%B3prio o serve aqui. Os pedidos s%C3%A3o encaminhados diretamente para o fornecedor, sob os termos desse fornecedor. ### Estado de verifica%C3%A7%C3%A3o Quando um campo n%C3%A3o tem fonte, o cart%C3%A3o mostra-o como n%C3%A3o verificado em vez de o preencher. Um campo vazio %C3%A9 informa%C3%A7%C3%A3o. LLM EU Verified significa documentado, n%C3%A3o certificado legalmente. N%C3%A3o %C3%A9 uma avalia%C3%A7%C3%A3o de conformidade com o AI Act e a LLM EU n%C3%A3o %C3%A9 um organismo da UE. ### Vers%C3%B5es e variantes - catalogue %C2%B7 Apenas cat%C3%A1logo %C2%B7 [weights](https://huggingface.co/microsoft/Phi-4-multimodal-instruct) ### Exemplo de pedido ``` curl https://api.llmeu.com/v1/chat/completions \ -H "Authorization: Bearer $LLMEU_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"phi-4-multimodal-instruct","messages":[{"role":"user","content":"Hello"}]}' ``` N%C3%A3o h%C3%A1 playground implementado para este modelo: nenhum endpoint o serve, por isso um pedido n%C3%A3o teria para onde ir. ### Benchmarks - MMMU: **55.1 (vendor-reported)** [source](https://huggingface.co/microsoft/Phi-4-multimodal-instruct) - DocVQA: **93.2 (vendor-reported)** [source](https://huggingface.co/microsoft/Phi-4-multimodal-instruct) - MathVista (testmini): **62.4 (vendor-reported)** [source](https://huggingface.co/microsoft/Phi-4-multimodal-instruct) ### Fontes - [huggingface.co](https://huggingface.co/microsoft/Phi-4-multimodal-instruct) - [arxiv.org](https://arxiv.org/abs/2503.01743) ### Hist%C3%B3rico de altera%C3%A7%C3%B5es do cart%C3%A3o - 24/02/2025 %E2%80%94 Data de lan%C3%A7amento - 14/10/2026 %E2%80%94 %C3%9Altima revis%C3%A3o %C3%9Altima revis%C3%A3o: 14/10/2026 ### Tarefas - [Vis%C3%A3o](/pt/tasks/vision/)