Spaces:
Sleeping
Sleeping
Commit History
-
feat(nebius-hf): Migrate default inference engine to Moonshot Kimi K2.5
This update sets Kimi K2.5 as the baseline multimodal and logic model, and Kimi K2 Thinking as the deep reasoning model, according to the 2026 SOTA benchmarks. Output token limits were also expanded to accommodate Reasoning tokens.eeba397