AMD's GAIA Defaults To Better Model, Continued Improvements For Local AI
([AMD] 101 Minutes Ago
Generative AI Is Awesome)
- Reference: 0001631392
- News link: https://www.phoronix.com/news/AMD-GAIA-0.17.5
- Source link:
AMD software engineers on Friday released a new version of GAIA "Generative AI Is Awesome" as their open-source software for Windows and Linux leveraging the Lemonade SDK and aiming to make it easy to build AI agents on your PC with all local AI processing across AMD's CPUs, GPUs, and NPUs.
AMD [1]GAIA continues steadily improving and making it easier to get local AI agents running on your AMD hardware whether it's on Windows or Linux. With GAIA 0.17.5, they have replaced Qwen 3.5 35B with Gemma 4 E4B as the default model. They explained their rationale in the release announcement as:
"Gemma 4 E4B (Gemma-4-E4B-it-GGUF) replaces Qwen 3.5 35B and the separate Qwen 3-VL-4B as the single default across the LLM and VLM roles, the installer profiles, the CLI, the Agent UI, and the eval suite). Gemma 4 is natively multimodal at ~4.5B effective parameters with a 128K context window and an Apache 2.0 licence, so one model now covers what previously required loading two. The post-swap eval baseline beats the pre-swap Qwen baseline 14/15 vs 13/15 across the bundled scenarios."
GAIA 0.17.5 also adds native OpenAI tool_calls path support, Chat Lite "chat-lite" is a new built-in agent with lighter resources for systems not able to handle the 35B chat default, and semantic code search via CodeAgent.
[2]
GAIA 0.17.5 also now has the agent UI bundled in the PyPi wheel. Thus a simple and easy install as pip install amd-gaia[ui] && gaia chat --ui for the real React-based UI.
[3]
When trying out GAIA 0.17.5 via PyPi, this time around was the smoothest I've had it up and running yet after a few tries in the past. It's becoming a nice and pleasant Linux experience with GAIA and Lemonade SDK.
[4]
But it's not yet perfect. I was encountering some errors with some of the default agents. While the Radeon 890M GPU was being utilized on the AMD Ryzen AI 9 HX 370 laptop for testing, the Ryzen AI NPU wasn't being utilized at all even with a fresh Ubuntu 26.04 base with AMDXDNA supported kernel and capable Lemonade 10.3 release. Across different agents and tasks, I never saw the AMD Ryzen AI NPU take off above 0% usage. But aside from that, the experience is getting there and closer to being a more compelling story for Linux AI usage on the desktop without any hair pulling.
Those wanting to learn more about the AMD GAIA 0.17.5 changes can do so via [5]GitHub .
[1] https://www.phoronix.com/search/Gaia
[2] https://www.phoronix.com/image-viewer.php?id=2026&image=amd_gaia_0175_1_lrg
[3] https://www.phoronix.com/image-viewer.php?id=2026&image=amd_gaia_0175_2_lrg
[4] https://www.phoronix.com/image-viewer.php?id=2026&image=amd_gaia_0175_3_lrg
[5] https://github.com/amd/gaia/releases/tag/v0.17.5
AMD [1]GAIA continues steadily improving and making it easier to get local AI agents running on your AMD hardware whether it's on Windows or Linux. With GAIA 0.17.5, they have replaced Qwen 3.5 35B with Gemma 4 E4B as the default model. They explained their rationale in the release announcement as:
"Gemma 4 E4B (Gemma-4-E4B-it-GGUF) replaces Qwen 3.5 35B and the separate Qwen 3-VL-4B as the single default across the LLM and VLM roles, the installer profiles, the CLI, the Agent UI, and the eval suite). Gemma 4 is natively multimodal at ~4.5B effective parameters with a 128K context window and an Apache 2.0 licence, so one model now covers what previously required loading two. The post-swap eval baseline beats the pre-swap Qwen baseline 14/15 vs 13/15 across the bundled scenarios."
GAIA 0.17.5 also adds native OpenAI tool_calls path support, Chat Lite "chat-lite" is a new built-in agent with lighter resources for systems not able to handle the 35B chat default, and semantic code search via CodeAgent.
[2]
GAIA 0.17.5 also now has the agent UI bundled in the PyPi wheel. Thus a simple and easy install as pip install amd-gaia[ui] && gaia chat --ui for the real React-based UI.
[3]
When trying out GAIA 0.17.5 via PyPi, this time around was the smoothest I've had it up and running yet after a few tries in the past. It's becoming a nice and pleasant Linux experience with GAIA and Lemonade SDK.
[4]
But it's not yet perfect. I was encountering some errors with some of the default agents. While the Radeon 890M GPU was being utilized on the AMD Ryzen AI 9 HX 370 laptop for testing, the Ryzen AI NPU wasn't being utilized at all even with a fresh Ubuntu 26.04 base with AMDXDNA supported kernel and capable Lemonade 10.3 release. Across different agents and tasks, I never saw the AMD Ryzen AI NPU take off above 0% usage. But aside from that, the experience is getting there and closer to being a more compelling story for Linux AI usage on the desktop without any hair pulling.
Those wanting to learn more about the AMD GAIA 0.17.5 changes can do so via [5]GitHub .
[1] https://www.phoronix.com/search/Gaia
[2] https://www.phoronix.com/image-viewer.php?id=2026&image=amd_gaia_0175_1_lrg
[3] https://www.phoronix.com/image-viewer.php?id=2026&image=amd_gaia_0175_2_lrg
[4] https://www.phoronix.com/image-viewer.php?id=2026&image=amd_gaia_0175_3_lrg
[5] https://github.com/amd/gaia/releases/tag/v0.17.5