What's new in RHEL 10.1: Offline assistance, convenient AI accelerators, and more

During the excitement of the Red Hat Enterprise Linux 10 (RHEL) launch at Red Hat Summit, I kept hearing one question from customers and partners: When would an offline version of the RHEL command-line assistant be available? Today I can announce that it's on the way.

As part of the RHEL 10.1 update, an offline, locally available command-line assistant is officially in developer preview. For customers with a Red Hat Satellite subscription, it offers AI-powered RHEL guidance based on decades of enterprise Linux experience. Companies and agencies in finance, government, defense, industrial control, and other heightened-security industries will find it particularly useful. A key advantage of this design is its ability to function in completely disconnected, offline, or air-gapped environments, eliminating the need for external network connectivity. This allows users to receive AI-powered guidance and suggestions for a wide range of Red Hat Enterprise Linux tasks, including questions related to the RHEL installation process, troubleshooting, and more, without compromising security or relying on cloud-based services.

Access to the offline, locally available command-line assistant requires a Red Hat Satellite subscription.

The RHEL command-line assistant gets another improvement in RHEL 10.1 (and RHEL 9.7, also launching today): an increase in its context limit, from 2KB to 32KB. Expanding the command-line assistant's working memory unlocks more powerful, elaborate interactions. Now it can take on larger log files, pipe extensive data streams, and help with bigger tasks.

AI accelerators close at hand

Recent RHEL updates have focused squarely on making it easier for customers to embrace AI. With RHEL 10.1, we're improving a key aspect of the interaction between the operating system and the hardware that supplies the raw computing power AI demands.

We're making vendor-validated AI accelerator drivers available in the RHEL Extensions Repository and Supplementary Repository. Like nearly everything in AI, hardware is advancing at a pace that data scientists and security teams struggle to maintain. Keeping current drivers for graphics processing units (GPU), tensor processing units (TPU),  and other app-specific integration circuits (ASIC) is difficult, and bringing vendor-verified, compatible, RHEL-ready drivers to an ecosystem IT teams already trust eliminates a lot of toil.

We'll start with drivers from three leading hardware makers:

• NVIDIA: OpenRM kernel mode driver, CUDA toolkit
• AMD: The amdgpu kernel mode driver and ROCm
• Intel: Neural processing unit (NPU) kernel mode driver

Image mode updates: Soft-reboots and reproducible builds

In RHEL 10.1, we're introducing a new systemd capability with image mode: Soft-reboots. This new systemd capability cuts downtime by letting administrators alter system state without fully rebooting. With this update, admins can update or reset applications, libraries and other userspace components without interrupting kernel operations. 

With a soft-reboot, customers using image mode can quickly apply security patches, update software, or reset system states with minimal service interruptions. More uptime means more agility for your organization.

One other efficiency improvement for image mode users: RHEL 10.1 and RHEL 9.7 both feature reproducible builds for container tools in image mode. Previously, container images based on the same inputs still had slight variations, such as different timestamps. With this update, container images created from matching content are truly identical, all the way down to their metadata. That's a boost for security, reliability, and efficiency.

Increasing developer productivity with updated toolsets

RHEL 10.1 includes updated versions of leading programming languages and services, including:

• Go 1.24 adds new standard library packages for weak pointers and crypto algorithms, support for generic type aliases, and several runtime performance improvements that reduce CPU overhead.
• LLVM 20 includes expanded hardware support, improvements to core libraries, and modernized just-in-time linking infrastructure, plus updates to the Clang and Flang tools.
• Rust 1.88 includes a stabilized Rust 2024 edition with significant language changes, and makes specific CPU features for high-performance computing accessible directly in safe Rust.
• GCC 15 supports increased program reliability with runtime assertions in the C++ standard library, now enabled by default for unoptimized builds. GCC Toolset 15 also includes a preview of the C++ standard library module
• .NET 10 features better runtime performance, new APIs for working with cryptography, globalization, numerics, collections, and ZIP files. It also features extended support for containers in the .NET SDK, and support for OpenAPI 3.1 in web applications.
• Valkey 8 brings intelligent multicore utilization and asynchronous I/O threading, improved cluster scaling with automatic failover for new shards and replicated migration states, faster replication with dual-channel relational database and replica backlog streaming, and better visibility through improved per-slot and per-client metrics.
• Node.js 24 features a new URLPattern added as a global object for further web compatibility, an update to the V8 JavaScript engine, and a promotion for the permission model from the experimental stage to production usage.

Fortifying a next-generation foundation

Post-quantum cryptography (PQC) was a leap forward for security in RHEL 10. In RHEL 10.1, we continue to evolve for a post-quantum world by enhancing PQC support for transport layer security (TLS). Introducing PQC algorithms to the transit layer shores up data security at points where information crosses networks, a gap for many organizations.

OpenTelemetry Collector, part of RHEL 9 and RHEL 10 cloud images, now includes support for Trusted Platform Module (TPM) on all three major public cloud platforms. By safeguarding cryptographic keys and attestation data in the cloud, TPM brings hardware-grade security to the software-only environments where hybrid IT increasingly works.

That applies to cloud virtual machine deployments, too. The OpenTelemetry Collector upgrade includes support for virtualized TPM (vTPM) for RHEL 10.1 and RHEL 9.7 cloud images, too.

Cloud-crossing consistency with simpler image creation

RHEL image builder has been a go-to tool for customers using RHEL in public clouds. In RHEL 10.1, the process of creating RHEL images gets simpler with the debut of our image builder command-line interface (CLI). With no need for continuously running services, image builder users can install and set up build environments more easily. That streamlines integration of RHEL image building into continuous integration/continuous delivery (CI/CD) pipelines and automated workflows.

Ultimately, it also lets organizations deploy with consistent settings and standardized processes, regardless of the environment. The image builder CLI is currently  in tech preview.

With RHEL 10.1, our operating system continues evolving so customers can operate consistently and embrace AI, PQC, and other new technologies as they emerge. Get more information about that evolution, and then test RHEL 10.1 yourself:

Additional resources:

• RHEL 10.1 Product Download
• RHEL 10.1 Release Notes
• RHEL 10.1 Documentation
• What’s New in RHEL 9.7 Blog
• RHEL 10.1/9.7 Press Release