webAI open-sources TwIL-8B, an 8.79B local reasoning model
webAI has published TwIL-8B weights on Hugging Face. The dense 8.79B reasoning model is post-trained from IBM Granite 4.2 8B for math, code and instruction following, and is meant to run on ordinary hardware; a 4-bit build is about 5 GB. The company reports a MATH-500 score of 98.2, ahead of several peers of similar size, while lagging on knowledge-heavy tests. It ships in balanced and verbose variants under a non-commercial license.
Why it matters
Another small open-weight model closing the gap with larger systems on math and code, useful for people who want capable local inference.
Sources1
Written by dalao.ai: the title, summary and why it matters are our own words. The original reports are the record; their publishers keep authorship.