#Pixtral Large 25.02 model

[govindhtech]

Pixtral Large 25.02: Amazon Bedrock Serverless Multimodal AI

AWS releases Pixtral Large 25.02 for serverless Amazon Bedrock.

Amazon Bedrock Pixtral Large

The Pixtral Large 25.02 model is now completely managed and serverless on Amazon Bedrock. AWS was the first major cloud service to provide serverless, fully managed Pixtral Large.

Infrastructure design, specific expertise, and continual optimisation are often needed to manage massive foundation model (FM) computational demands. Many clients must manage complex infrastructures or choose between cost and performance when deploying sophisticated models.

Mistral AI's first multimodal model, Pixtral Large, combines high language understanding with advanced visuals. Its 128K context window makes it ideal for complex visual reasoning. The model performs well on MathVista, DocVQA, and VQAv2, proving its effectiveness in document analysis, chart interpretation, and natural picture understanding.

Pixtral Large excels at multilingualism. Global teams and apps can use English, French, German, Spanish, Italian, Chinese, Japanese, Korean, Portuguese, Dutch, and Polish. Python, Java, C, C++, JavaScript, Bash, Swift, and Fortran are among the 80 languages it can write and read.

Developers will like the model's agent-centric architecture since it integrates with current systems via function calling and JSON output formatting. Its sturdy system fast adherence improves dependability in large context situations and RAG applications.

This complex model is currently available in Amazon Bedrock without infrastructure for Pixtral Large. Serverless allows you to scale usage based on demand without prior commitments or capacity planning. No wasted resources mean you only pay for what you utilise.

Deduction across regions

Pixtral Large is now available in Amazon Bedrock across various AWS Regions due to cross-region inference.

Amazon Bedrock cross-Region inference lets you access a single FM across many regions with high availability and low latency for global applications. A model deployed in both the US and Europe may be accessible via region-specific API endpoints with various prefixes: us.model-id for US and eu.model-id for European.

By confining data processing within defined geographic borders, Amazon Bedrock may comply with laws and save latency by sending inference requests to the user's nearest endpoint. The system automatically manages load balancing and traffic routing across Regional installations to enable seamless scalability and redundancy without your monitoring.

How it works?

I always investigate how new capabilities might solve actual problems as a developer advocate. The Amazon Bedrock Converse API's new multimodal features were perfect for testing when she sought for help with her physics exam.

It struggled to solve these challenges. It realised this was the best usage for the new multimodal characteristics. The Converse API was used to create a rudimentary application that could comprehend photographs of a complex problem sheet with graphs and mathematical symbols. Once the physics test materials were uploaded, ask the model to explain the response process.

The following event impressed them both. Model interpreted schematics, mathematical notation, and French language, and described each difficulty step-by-step. The computer kept context throughout the talk and offered follow-up questions about certain issues to make teaching feel natural.

It was confident and ready for this test, showing how Amazon Bedrock's multimodal capabilities can provide users meaningful experiences.

Start now

The new method is available at US East (Ohio, N. Virginia), US West (Oregon), and Europe (Frankfurt, Ireland, Paris, Stockholm) Regional API endpoints. Regional availability reduces latency, meeting data residency criteria.

Use the AWS CLI, SDK, or Management Console to programmatically access the model using the model ID mistral.pixtral-large-2502-v1:0.

Developers and organisations of all sizes may now employ strong multimodal AI, a major leap. AWS serverless infrastructure with Mistral AI's cutting-edge model let you focus on designing innovative apps without worrying about complexity.

[jcmarchi]

When AI Backfires: Enkrypt AI Report Exposes Dangerous Vulnerabilities in Multimodal Models

New Post has been published on https://thedigitalinsider.com/when-ai-backfires-enkrypt-ai-report-exposes-dangerous-vulnerabilities-in-multimodal-models/

When AI Backfires: Enkrypt AI Report Exposes Dangerous Vulnerabilities in Multimodal Models

In May 2025, Enkrypt AI released its Multimodal Red Teaming Report, a chilling analysis that revealed just how easily advanced AI systems can be manipulated into generating dangerous and unethical content. The report focuses on two of Mistral’s leading vision-language models—Pixtral-Large (25.02) and Pixtral-12b—and paints a picture of models that are not only technically impressive but disturbingly vulnerable.

Vision-language models (VLMs) like Pixtral are built to interpret both visual and textual inputs, allowing them to respond intelligently to complex, real-world prompts. But this capability comes with increased risk. Unlike traditional language models that only process text, VLMs can be influenced by the interplay between images and words, opening new doors for adversarial attacks. Enkrypt AI’s testing shows how easily these doors can be pried open.

Alarming Test Results: CSEM and CBRN Failures

The team behind the report used sophisticated red teaming methods—a form of adversarial evaluation designed to mimic real-world threats. These tests employed tactics like jailbreaking (prompting the model with carefully crafted queries to bypass safety filters), image-based deception, and context manipulation. Alarmingly, 68% of these adversarial prompts elicited harmful responses across the two Pixtral models, including content that related to grooming, exploitation, and even chemical weapons design.

One of the most striking revelations involves child sexual exploitation material (CSEM). The report found that Mistral’s models were 60 times more likely to produce CSEM-related content compared to industry benchmarks like GPT-4o and Claude 3.7 Sonnet. In test cases, models responded to disguised grooming prompts with structured, multi-paragraph content explaining how to manipulate minors—wrapped in disingenuous disclaimers like “for educational awareness only.” The models weren’t simply failing to reject harmful queries—they were completing them in detail.

Equally disturbing were the results in the CBRN (Chemical, Biological, Radiological, and Nuclear) risk category. When prompted with a request on how to modify the VX nerve agent—a chemical weapon—the models offered shockingly specific ideas for increasing its persistence in the environment. They described, in redacted but clearly technical detail, methods like encapsulation, environmental shielding, and controlled release systems.

These failures were not always triggered by overtly harmful requests. One tactic involved uploading an image of a blank numbered list and asking the model to “fill in the details.” This simple, seemingly innocuous prompt led to the generation of unethical and illegal instructions. The fusion of visual and textual manipulation proved especially dangerous—highlighting a unique challenge posed by multimodal AI.

Why Vision-Language Models Pose New Security Challenges

At the heart of these risks lies the technical complexity of vision-language models. These systems don’t just parse language—they synthesize meaning across formats, which means they must interpret image content, understand text context, and respond accordingly. This interaction introduces new vectors for exploitation. A model might correctly reject a harmful text prompt alone, but when paired with a suggestive image or ambiguous context, it may generate dangerous output.

Enkrypt AI’s red teaming uncovered how cross-modal injection attacks—where subtle cues in one modality influence the output of another—can completely bypass standard safety mechanisms. These failures demonstrate that traditional content moderation techniques, built for single-modality systems, are not enough for today’s VLMs.

The report also details how the Pixtral models were accessed: Pixtral-Large through AWS Bedrock and Pixtral-12b via the Mistral platform. This real-world deployment context further emphasizes the urgency of these findings. These models are not confined to labs—they are available through mainstream cloud platforms and could easily be integrated into consumer or enterprise products.

What Must Be Done: A Blueprint for Safer AI

To its credit, Enkrypt AI does more than highlight the problems—it offers a path forward. The report outlines a comprehensive mitigation strategy, starting with safety alignment training. This involves retraining the model using its own red teaming data to reduce susceptibility to harmful prompts. Techniques like Direct Preference Optimization (DPO) are recommended to fine-tune model responses away from risky outputs.

It also stresses the importance of context-aware guardrails—dynamic filters that can interpret and block harmful queries in real time, taking into account the full context of multimodal input. In addition, the use of Model Risk Cards is proposed as a transparency measure, helping stakeholders understand the model’s limitations and known failure cases.

Perhaps the most critical recommendation is to treat red teaming as an ongoing process, not a one-time test. As models evolve, so do attack strategies. Only continuous evaluation and active monitoring can ensure long-term reliability, especially when models are deployed in sensitive sectors like healthcare, education, or defense.

The Multimodal Red Teaming Report from Enkrypt AI is a clear signal to the AI industry: multimodal power comes with multimodal responsibility. These models represent a leap forward in capability, but they also require a leap in how we think about safety, security, and ethical deployment. Left unchecked, they don’t just risk failure—they risk real-world harm.

For anyone working on or deploying large-scale AI, this report is not just a warning. It’s a playbook. And it couldn’t have come at a more urgent time.