Article · MarkTechPost ·
webAI released TwIL-8B, an 8.8B reasoning model post-trained from IBM Granite 4.2 8B. It posts 98.2 on MATH-500 and 81.6 on LiveCodeBench, but trails similar-size models on knowledge benchmarks.
ArticleMarkTechPost
Ricursive Intelligence released Ripple-65B Preview, a 65B dense open-weight LLM under Apache 2.0. It scores 80.5 on Terminal-Bench 2.1, ahead of US open models up to 15x its size.
ArticleMarkTechPost
Tencent’s WeChat Vision team released WeVisOne-VL, an Apache 2.0 vision language model in 2B and 4B sizes. It handles detection, OCR, counting, segmentation and alpha matting in one model, with a 9.6M-parameter pixel decoder.
ArticleMarkTechPost3 sources
Anthropic released FlashProve, an Apache 2.0 zkSNARK prototype that proves a reply came from a committed model without revealing weights. It proves Llama 4 Maverick at 110 tokens/second on 1 H100, with 19 MB proofs.
ArticleMarkTechPost
NVIDIA shows how 3 stacked optimizations speed up MoE inference on a single DGX Spark. On Ant Group inclusionAI's 124B Ling-3.0-flash, GSM8K decode throughput rose 42.5% with no detectable accuracy change.