View this package for all CentOS/RHEL versions →

How to install ollama in CentOS/RHEL 9 (aarch64)

Get up and running with large language models locally

License: MIT https://github.com/ollama/ollama

Install

sudo dnf -y install https://extras.getpagespeed.com/release-latest.rpm
sudo dnf -y install ollama

Description

Ollama allows you to run large language models locally. Get up and running with Llama 3, Mistral, Gemma, and other open source models.

Architecture

Optimized for aarch64 (ARM64) architecture. It can be used on AWS Graviton instances, as well as Raspberry Pi 4 and newer.

Compatibility

This package is compatible with all RHEL-based distributions, including CentOS, AlmaLinux, Oracle Linux, Rocky Linux, etc.

RPMs

Package	Size	Changelog
ollama-0.21.2-1.el9.aarch64	10.4 MiB	Danila Vershinin (2026-04-24) - Hardened OpenClaw onboarding flow for improved security. · - Enhanced user experience during onboarding process.
ollama-0.21.1-1.el9.aarch64	10.4 MiB	Danila Vershinin (2026-04-22) - Updated top recommended model to k2.6 from kimi-k2.5. · - Improved model performance and compatibility with latest features.
ollama-0.21.0-1.el9.aarch64	10.4 MiB	Danila Vershinin (2026-04-18) - Added Hermes agent for automated skill creation. · - Supported Gemma 4 on MLX for Apple Silicon. · - Integrated GitHub Copilot CLI with ollama launch. · - OpenCode now uses inline configuration. · - ollama launch no longer rewrites unchanged config. · - Fixed non-interactive setup for openclaw command. · - Resolved Gemma 4 compiler error affecting Metal builds.
ollama-0.20.7-1.el9.aarch64	10.3 MiB	Danila Vershinin (2026-04-14) - Improved quality for gemma:e2b and gemma:e4b with thinking disabled. · - Updated ROCm to version 7.2.1 for Linux.
ollama-0.20.6-1.el9.aarch64	10.3 MiB	Danila Vershinin (2026-04-13) - Improved tool calling with Gemma 4. · - Fixed various application bugs. · - Enhanced parallel tool calling performance. · - Updated documentation for Hermes Agent.
ollama-0.20.5-1.el9.aarch64	10.3 MiB	Danila Vershinin (2026-04-11) - Added OpenClaw channel setup for messaging apps. · - Enabled flash attention for Gemma 4 on compatible GPUs. · - Improved detection of curl-based OpenCode installs. · - Fixed /save command for safetensors-based models.
ollama-0.20.4-1.el9.aarch64	10.3 MiB	Danila Vershinin (2026-04-09) - Improved M5 performance with NAX. · - Enabled flash attention for gemma4.
ollama-0.20.3-1.el9.aarch64	10.3 MiB	Danila Vershinin (2026-04-08) - Improved Gemma 4 Tool Calling functionality. · - Added latest models to Ollama App. · - Fixed OpenClaw issues for TUI launching.
ollama-0.20.2-1.el9.aarch64	10.3 MiB	Danila Vershinin (2026-04-05) - Default app home view set to new chat. · - Improved user experience on application launch.
ollama-0.20.0-1.el9.aarch64	10.3 MiB	Danila Vershinin (2026-04-03) - Added support for Effective 2B and 4B models. · - Introduced 26B and 31B model options. · - Updated documentation for PI features. · - Tokenizer now respects add_bos_token setting. · - Added SentencePiece-style BPE support.