bibibi.ccLoading…
Connect scattered updates into a story you can understand.
PrismML has added to Bonsai 2 three new 2B、4B、9B models, saying they are based on Qwen3.5 and use ternary weights to compress the file size. 9B The 18 model reportedly achieves up to 34 Token per second on the iPhone
The report says Apple open-sourced a Qwen3.5-9B-Base trained on LensVLM-9B, which processes long documents by filtering with low-resolution pages before reading the original text. It says this can significantly reduce context and KV cache usage, but each answer takes longer.
Belgian cybersecurity company Aikido released Altar-1, compressing GLM-5.3 to 328GB through quantization and expert pruning for local code auditing and penetration testing.