KIEFERSA/Sophea-Titan-1

DeepSeek-V4.1-Flash: add MMStar vision result (0.721 overall, thinking off, letter-answer)

Add DeepSeek-V4.1-Flash column to the Frontier & other API models tables (thinking off, greedy, 500 q/bench; Greek macro 0.845, English macro 0.888)

Add Qwen3.8-Max column to Frontier & other API models tables (Qwen Cloud API, thinking disabled, temperature=0; Greek macro 89.84, English macro 93.48)

Add Kimi-K3 (Moonshot API, thinking-on) to Frontier & other API models tables — Greek macro 0.9158, English macro 0.9472

Add Inkling-Small (NVFP4, thinking-on) to Frontier & other API models tables — Greek macro 0.881, English macro 0.922

Restore the MTP head dropped by the LoRA merge

Sophea-Titan-1 (dense) release