OpenAI releases research preview GPT-5.3-Codex-Spark for ChatGPT Pro users
GPT-5.3-Codex-Spark is a lightweight version of the company’s coding model, GPT-5.3-Codex, that is optimized to run on ultra-low latency hardware and can deliver over 1,000 tokens per second.
It is the first outcome of the company’s recently announced partnership with Cerebras to add 750MW of ultra low-latency AI compute. OpenAI says the reason it is being released as a research preview is to provide developers with the opportunity to start benefiting from this partnership while the company continues ramping up Cerabras in its data centers.
“Codex-Spark is our first model designed specifically for working with Codex in real-time—making targeted edits, reshaping logic, or refining interfaces and seeing results immediately. With Codex-Spark, Codex now supports both long-running, ambitious tasks and getting work done in the moment. We hope to learn from how developers use it and incorporate feedback as we continue to expand access,” OpenAI wrote in a post.
GitHub starts technical preview for Agentic Workflows
GitHub Agentic Workflows allow developers to describe the outcome they want in plain Markdown, add it as an automated workflow to their repository, and have that be executed as a coding agent in GitHub Actions.
Agentic Workflows run as standard GitHub Actions workflows that can have additional guardrails for sandboxing, permissions, control, and review. Additionally, they support a variety of coding agent engines, including Copilot CLI, Claude Code, or OpenAI Codex.
“The use of GitHub Agentic Workflows makes entirely new categories of repository automation and software engineering possible, in a way that fits naturally with how developer teams already work on GitHub,” GitHub wrote in a post.
Google adds Automated Reviews to Conductor in the Gemini CLI
Conductor is a Gemini CLI extension that helps bring additional development context into the terminal, and the new Automated Review feature generates a post-implementation report after the agent completes its coding tasks. Findings are categorized by severity (low, medium, and high) so that developers can prioritize where to iterate first.
“This level of detail ensures that “agentic” development doesn’t mean ‘unsupervised’ development. Instead, it creates a workflow where the AI provides the labor and the developer provides the high-level architectural oversight, backed by automated verification,” Google wrote in a blog post.
Additionally, Google announced that Gemini CLI extensions will now be able to define settings that the user will be prompted to provide when installing an extension. By providing things like API keys, base URLs, and project identifiers upfront, users will hopefully have fewer configuration errors to troubleshoot when working with the Gemini CLI, the company explained.
Google upgrades Gemini 3 Deep Think mode
According to Google, the upgraded model features improvements across math and programming reasoning, as well as specific scientific domains like chemistry and physics. It available for Google AI Ultra subscribers in the Gemini app, and for select researchers, engineers, and enterprises in the Gemini API for the first time.
“We updated Gemini 3 Deep Think in close partnership with scientists and researchers to tackle tough research challenges — where problems often lack clear guardrails or a single correct solution and data is often messy or incomplete. By blending deep scientific knowledge with everyday engineering utility, Deep Think moves beyond abstract theory to drive practical applications,” Google wrote in a post.
GitHub Copilot testing for .NET is now generally available
Now available in Visual Studio 2026 v18.3, the testing feature allows developers to generate unit tests within their IDE. The company added new capabilities to coincide with this GA release, including deeper IDE integration, more natural prompting, and new ways to invoke the testing experience.
The team plans to focus next on getting it to handle more advanced testing requests, which will involve addressing requirements like allowing developers to clarify intent, confirm assumptions, and review proposed plans before generating tests.
“General availability is an important milestone, but it is not the end of the journey. We continue to run user studies and gather feedback to understand how developers use GitHub Copilot testing for .NET in real-world scenarios, especially as requests grow in size and complexity,” Microsoft wrote in a blog post.
Anthropic raises $30 billion in Series G funding
This latest funding round was led by GIC and Coatue; co-led by D. E. Shaw Ventures, Dragoneer, Founders Fund, ICONIQ, and MGX; and had participation from 30 other investors. As of this round, Anthropic is now valued at $380 billion post-money.
In its announcement, Anthropic also revealed that its run-rate revenue is now $14 billion, and it has grown by 10x every year since its first funding round three years ago.
“Whether it is entrepreneurs, startups, or the world’s largest enterprises, the message from our customers is the same: Claude is increasingly becoming critical to how businesses work,” said Krishna Rao, chief financial officer of Anthropic. “This fundraising reflects the incredible demand we are seeing from these customers, and we will use this investment to continue building the enterprise-grade products and models they have come to depend on.”
