Muse Spark 1.3 & Gemini 3.8 Flash: Gemini has leveled up BIG TIME!
Visit Verdent: https://www.verdent.ai/?id=700712 Visit AISeeKing (my second channel with more cool ai stuff): https://www.youtube.com/@aiseeking In this video, I’ll be testing Meta’s new Muse Spark 1.3 and Google’s Gemini 3.8 Flash on KingBench 3. I’ll compare their frontend, 3D, SVG, maths, and agentic coding performance, while also discussing real-world usage, file-overwriting issues, and where you can try both models. -- Key Takeaways: 🚀 Gemini 3.8 Flash scored 81.25% on KingBench 3, placing it among the top five models on the leaderboard. 📈 Gemini 3.8 Flash delivers a massive improvement over Gemini 3.5 Flash, which scored only 30%. 📉 Muse Spark 1.3 scored 71.25%, falling below Muse Spark 1.2’s previous score of 76.25%. ⌚ Muse Spark 1.3 achieved the highest score yet on the difficult 3D wristwatch test. 🎨 Muse Spark regressed significantly in SVG generation and the Three.js folding-table challenge. 🏹 Gemini 3.8 Flash earned a perfect score on the bow-and-arrow simulator and performed strongly in visual tasks. 🤖 Both models completed the full agentic fine-tuning task and correctly solved the difficult maths problem. ⚠️ Both models tend to overwrite entire files instead of making targeted edits, so carefully reviewing diffs is important. 🛠️ Gemini 3.8 Flash is available through Antigravity, while both models can also be used in Verdent. 👍 Overall, Gemini 3.8 Flash is an impressive fast and affordable model, while Muse Spark 1.3 feels like a disappointing regression.
