Bijan Bowen tests Kimi K3 acrossEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. a browser operating system, a C++ skate game, interactive product sites, CAD-style modeling, 3D environments and a long-running multi-agent game buildA multi-agent system contains multiple AI agents that interact, coordinate, divide work, or influence one another while pursuing tasks.. The model frequently produces detailed, interactive results that approach proprietary frontier systems.
Bijan Bowen praises Kimi K3's visual quality, persistence and broad competence, especially in complex 3D scenes3D scene generation uses AI or procedural tools to create a digital three-dimensional environment from instructions, references, or structured inputs. and multi-step builds. The main weaknesses are excessive reasoning timeReasoning effort is the amount of internal computational work an AI model applies before producing an answer or action., occasional control or geometry failures and runs that can consume hours before revealing a blocking bug.
Bijan Bowen concludes that Kimi K3 is the best open-weight modelAn open-weight model makes its trained parameter values available so others can run or adapt it under the release's licence. he has tested and can credibly enter comparisons with leading closed models. Its much larger scale and higher API price make it less accessible than earlier Kimi releases, but the capability jump is substantial.
Watch on YouTube




