Ornith-1.5 Local AI - Better than Qwen 3.8 Max ALREADY? 🤯

Value of the Information & Strength of the Argument

The video’s primary value lies in its concrete demonstrations of the model’s capabilities and limitations, offering viewers a realistic impression of what Ornith-1.5 can do in practice. The creator backs each claim with live examples, showing both successes and failures, which is more informative than just citing benchmark scores. The argumentation is persuasive in the sense that it relies on empirical evidence rather than speculation. However, the evaluation is anecdotal, based on a single machine and specific prompts, and the comparisons to other models are not controlled. The creator also injects humor and personal opinion, which, while engaging, can bias the interpretation. Overall, the video provides a useful qualitative snapshot but does not constitute a rigorous scientific comparison.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is moderate. The creator does not reference external studies or peer-reviewed sources, relying primarily on the model’s Hugging Face page and his own testing. The sources cited (Hugging Face links, companion videos) are relevant and provide context, but there is no critical analysis of the benchmark claims. The title is somewhat sensationalist, implying a definitive ranking, but the content reveals a nuanced view, aligning with a 3-star rating. The adequacy of the title is acceptable but slightly overstated; the video does not conclusively prove that Ornith-1.5 is better, only that it performs competitively in some tests. The creator’s transparency about the testing setup (e.g., M3 Ultra, memory usage, token counts) adds credibility, but the lack of a systematic methodology reduces the overall reliability.

The title asks whether Ornith-1.5 is already better than Qwen 3.8 Max; the video attempts to answer through practical tests, but the conclusion is mixed and far from definitive.

6/10

The video provides a hands-on, qualitative evaluation of the Ornith-1.5 model across various tasks, with real outputs and some quantitative details (token counts, memory usage). However, the assessment is subjective, lacks a controlled benchmark protocol, and is limited to the creator's specific testing environment.

• 00:00Intro and Multi-modal Vision Capabilities
• 01:27Logic and Reasoning: Mathematics & Oranges Tests
• 03:32Advanced Logic: B-Selector and Car Wash Challenges
• 05:243D Simulations: Solar System and Voxel Art
• 07:02Coding Challenge: 3D Flight Simulator
• 08:25Scientific Rendering: Human Anatomy and Physics
• 10:48Game Development: GTA Clone & Thinking Enabled
• 13:25Classic Game Remakes: Red Dead and Mario
• 16:29Racing Simulations and Canvas Animation
• 18:02Productivity Apps and Final Logic Summary

• Inferencer App — The video uses this app to run the model locally.
• Ornith-1.5-397B-MLX-Q9 - Hugging Face — The official model card for the tested version.
• Inferencer Labs on Hugging Face — Organization page with related models like Kimi K3.
• Qwen 3.8 Review (Companion Video) — Referenced for comparison with Ornith-1.5.
• Kimi K3 Review (Companion Video) — Referenced for comparison with Ornith-1.5.
• GLM 5.2 Review (Companion Video) — Referenced as a model Ornith-1.5 claims to beat.

• Ornith-1.5 Model Card — Provides official benchmarks and technical details that align with some of the video's observations.

• GLM 5.2 Review (Companion Video) — The companion video reviews GLM 5.2, which the Ornith-1.5 model claims to beat, but the practical tests in this video do not show a clear superiority, creating a discrepancy.

• vtudio.com