Anthropic releases Claude Sonnet 4.6
Claude Sonnet 4.6 features improved skills in coding, computer use, long-context reasoning, agent planning, knowledge work, and design.
It is now the default model in claude.ai and Claude Cowork, has a 1M context window (beta), and is priced the same as Sonnet 4.5, at $3 per million input tokens and $15 per million output tokens.
“Performance that would have previously required reaching for an Opus-class model—including on real-world, economically valuable office tasks—is now available with Sonnet 4.6. The model also shows a major improvement in computer use skills compared to prior Sonnet models,” Anthropic wrote in a post.
Gemini 3.1 Pro now available in preview
Gemini 3.1 Pro is now available for developers in the Gemini API in Google AI Studio, Gemini CLI, Google Antigravity, and Android Studio. It can also be accessed in Vertex AI, Gemini Enterprise, the Gemini app, and NotebookLM.
“Building on the Gemini 3 series, 3.1 Pro represents a step forward in core reasoning. 3.1 Pro is a smarter, more capable baseline for complex problem-solving. This is reflected in our progress on rigorous benchmarks. On ARC-AGI-2, a benchmark that evaluates a model’s ability to solve entirely new logic patterns, 3.1 Pro achieved a verified score of 77.1%. This is more than double the reasoning performance of 3 Pro,” Google wrote in a post.
OpenAI adds Lockdown Mode, Elevated Risk labels to ChatGPT
These new features are designed to reduce the risk of prompt injection attacks.
Lockdown Mode restricts how ChatGPT is able to interact with external systems, reducing the chance of data exfiltration from a prompt injection attack, while the new Elevated Risk labels will be displayed on certain products to inform users that interacting with a specific feature may introduce additional risk. For example, developers can grant Codex network access so that it can do things like look up documentation online, but this extra access can also be risky. For now, Elevated Risk labels will be displayed in ChatGPT, ChatGPT Atlas, and Codex.
Microsoft creates a suite of pre-built agents for Visual Studio
The pre-built agents include Debugger, which uses call stacks, variable state, and diagnostic tools to work through errors; Profiler, which identifies bottlenecks and suggests optimizations; Test, which generates unit tests; and Modernize, which executes framework and dependency upgrades.
“Each preset agent is designed around a specific developer workflow and integrates with Visual Studio’s native tooling in ways that a generic assistant can’t,” Microsoft wrote in a blog post.
Agents can be accessed through the chat panel by using the agent picker or “@”.
GraphRAG enables more context-aware and verifiable responses from LLMs
Graphwise’s new GraphRAG offering acts as a semantic layer on top of knowledge graphs that LLMs can utilize to provide context-rich and verifiable answers.
According to the company, a typical RAG implementation flattens data into chunks, and with that approach, it can find similar words, but isn’t able to understand complex relationships, hierarchies, or logic connecting business data. On top of that, it is also usually difficult to see how an LLM came to its answer and what sources it used.
Graphwise believes that GraphRAG solves these issues by providing a pipeline where every step can be inspected and answers are backed by documents and graph entities.
It leverages several different search approaches, including retrieval from a knowledge graph, vector search in a specified vector store, and full-text search to enable keyword-driven discovery. It utilizes a knowledge-model-driven input processing approach to understand the user’s intent, allowing it to enrich concepts using the company’s taxonomy or ontology, expand queries using related entities and terms, and build a graph representation of the question.
Checkmarx enhances IDE-native agentic application security in Kiro
Agentic AI security provider Checkmarx announced an integration with the AWS Kiro IDE to enable developers working in that platform to identify and deal with security issues as code is written, the company said.
The integration puts Checkmarx Developer Assist directly into Kiro, so developers don’t have to leave the IDE to analyze the code for security.
Once developers activate Developer Assist inside Kiro and it is authenticated, Checkmarx said the tool will analyze source code and dependencies in the active workspace. Further, it said the tool will automatically surface security findings in the IDE, along with contextual data that helps developers fix security issues early in the development cycle. That data can be viewed in the Checkmarx One platform, providing stakeholders with a view of project risks.
Quest Trusted Data Management Platform makes it easier for organizations to create reusable data products
The Quest Trusted Data Management Platform unifies data modeling, data cataloging, data governance, data quality, and a data marketplace to enable organizations to deliver AI-ready data throughout their business.
“Building trusted AI-ready data and reusable data products can take up to six months, but your business can’t afford to wait, so teams skip the metadata, bypass governance workflows, and ignore data quality, and every department ends up with their own version of a data product. That results in fragmented, siloed data that isn’t trustworthy,” Quest Software explained in a video.
One of the key capabilities of the platform is the Automated Data Product Factory, which uses generative AI to create data products from natural language prompts, reducing data product design cycles, lowering delivery costs, and enabling business users to create their own data products.