Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: A teenage Chinese exchange student in an 80s suburban front yard at dusk wearing a creative Halloween costume that blends traditional Chinese elements with American trick-or-treat style, standing confidently among fallen autumn leaves with decorated houses and glowing jack-o-lanterns in the background, warm nostalgic lighting, cinematic 1980s film aesthetic
MiniMax M2 + Claude Code on KingBench Agentic Evaluations: It now scores #2 on my Agentic Evaluations beating GLM-4.6 by a wide margin. It seems to work much better with Claude Code’s Tools. Really great model and it’s my daily driver now. I haven’t tested GLM with CC yet. https://x.com/aicodeking/status/1983934597353402797
A ready-to-run demo script is provided to compare a baseline text model (e.g., Qwen3, GLM4) with Glyph for long-context inference efficiency: https://x.com/Zai_org/status/1982804378667888808
Glyph can achieve 3-4x token compression on long text sequences without performance degradation, enabling significant gains in memory efficiency, training throughput, and inference speed. https://x.com/Zai_org/status/1982804372489646586
Glyph: Scaling Context Windows via Visual-Text Compression Paper: https://x.com/Zai_org/status/1982804366475063446
