muse-spark-1.2-contributor seems really bad at following Chinese prompts?
Compared notes with a few people, same impression across the board. It keeps misreading what you ask in Chinese. Some say Meta is quietly nerfing the version it serves Chinese users. Maybe. I think the model's just weak in Chinese.
muse-spark-1.2-contributor seems really bad at following Chinese
More posts
Opencode killed the $5 first-month promo on its Go plan
. First month is $10 flat now. Their cost pressure must be real — they also just dropped the free DS-V4-Flash-Free model.
Marketing genius over at Cursor, honestly.
They're running a promo: cancel your other AI subs and you get a free month of Grok Bot (the $200 plan). 🔗 anysphere.typeform.com/grokbot
Meituan shipped LongCat-2.5
. 1.6T params, ~48B active, native multimodal, aimed at long-horizon agents, trained start to finish on Chinese hardware. Images get quantized into multi-level discrete tokens, and that's where Visual Coding and 3D pick up. On the RL side, failed trajectories get turned back into verifiable training tasks. W4A4 inference, 43% faster prefill. Meituan's own evals put it between Claude Opus 4.8 and Opus 5.
SenseTime open-sourced SenseNova Loop-9B-Preview.
It stacks layers instead of parameters. The decoder has 32 layers, but layers 4 through 27 run twice with shared weights, so a forward pass does 56 layers of work while the checkpoint stays at 9B. Their own eval numbers: 71.38 on BIRD-Dev and 13.72 on AutomationBench, both above Qwen3.5-27B. Terminal Bench is still behind.
