--- title: webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring... description: webAI released TwIL-8B, an 8.79B parameter reasoning model post-trained from IBM's Granite 4.2 8B for math, code and instruction following. It scores 98.2 on... canonical: https://daily.dev/posts/webai-releases-twil-8b-an-8-8b-reasoning-model-scoring-98-2-on-math-500-ymqq9zelv twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500 | daily.dev og:description: webAI released TwIL-8B, an 8.79B parameter reasoning model post-trained from IBM's Granite 4.2 8B for math, code and instruction following. It scores 98.2 on... og:url: https://daily.dev/posts/webai-releases-twil-8b-an-8-8b-reasoning-model-scoring-98-2-on-math-500-ymqq9zelv og:image: https://api.daily.dev/og/posts/ymQq9ZElv.png og:image:alt: webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500 og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500 **[Machine Learning News](https://daily.dev/sources/mlnews)** · 5 min read · 0 upvotes · 0 comments ## Summary webAI released TwIL-8B, an 8.79B parameter reasoning model post-trained from IBM's Granite 4.2 8B for math, code and instruction following. It scores 98.2 on MATH-500, the top result in webAI's comparison, and 81.6 on LiveCodeBench for the verbose variant. It trails on knowledge benchmarks, scoring 70.2 on GPQA Diamond versus 81.7 for Qwen3.5-9B. The model uses 38% fewer output tokens than Qwen3.5-9B on AIME 2026. It runs locally via a 4.98 GiB Q4_K_M GGUF build, with weights available on Hugging Face under a non-commercial license. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How does webAI's TwIL-8B compare to Qwen3.5-9B on math and code benchmarks? TwIL-8B scores higher on math and code but lower on knowledge tasks. It posts 98.2 on MATH-500 versus 96.6 for Qwen3.5-9B, and 81.6 on LiveCodeBench (verbose variant) versus 65.6. On AIME 2026 Qwen3.5-9B leads 92.5 to 88.3. Across 13 shared benchmarks, TwIL-8B wins 7-6 overall. _Comparing small reasoning models before picking one? Track benchmark releases like this on daily.dev._ ### How much GPU memory or disk space does TwIL-8B need to run locally? The Q4_K_M GGUF build is 4.98 GiB and runs on CPU or a modest-VRAM GPU via llama.cpp, suitable for both the balanced and verbose variants. Full bf16 weights total 16.38 GiB, and a Q8_0 build is 8.70 GiB for the verbose variant only. All reported benchmark scores used bf16 weights, so quantized runs may score slightly lower. _Planning local LLM deployments? Daily.dev surfaces model release specs like these as they ship._ ### Is TwIL-8B licensed for commercial use? No, TwIL-8B's weights are open on Hugging Face but licensed for non-commercial use only. The model is an 8.79B parameter dense reasoning model post-trained from IBM's Granite 4.2 8B, with no hosted inference provider serving it yet. _Vetting model licenses before adoption? daily.dev tracks open-weight release terms as they land._ ## Similar posts on daily.dev - [ThursdAI - Jul 16 - Inkling 975B open weights, Kimi K3 at 2.8T, a 27B model on a phone & Codex hits 9M](https://daily.dev/posts/thursdai---jul-16---inkling-975b-open-weights-kimi-k3-at-2-8t-a-27b-model-on-a-phone-codex-hits--hledq9pdh) · ThursdAI · 0 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#open-source](https://daily.dev/tags/open-source), [#local-ai](https://daily.dev/tags/local-ai) [View this post on daily.dev](https://daily.dev/posts/webai-releases-twil-8b-an-8-8b-reasoning-model-scoring-98-2-on-math-500-ymqq9zelv) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500","url":"https://daily.dev/posts/webai-releases-twil-8b-an-8-8b-reasoning-model-scoring-98-2-on-math-500-ymqq9zelv","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/webai-releases-twil-8b-an-8-8b-reasoning-model-scoring-98-2-on-math-500-ymqq9zelv"},"datePublished":"2026-11-04T22:54:59.182Z","dateModified":"2026-11-05T06:30:49.856Z","description":"webAI released TwIL-8B, an 8.79B parameter reasoning model post-trained from IBM's Granite 4.2 8B for math, code and instruction following. It scores 98.2 on...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/69d77d1d6b461e41777cc33713d98760?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/69d77d1d6b461e41777cc33713d98760?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Machine Learning News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Machine Learning News","logo":"https://media.daily.dev/image/upload/s--0PoEtdkd--/f_auto/v1750409401/squads/marktechpost?_a=BAMClqZW0","url":"https://daily.dev/squads/mlnews"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/webai-releases-twil-8b-an-8-8b-reasoning-model-scoring-98-2-on-math-500-ymqq9zelv","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,open-source,local-ai","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Machine Learning News","item":"https://daily.dev/squads/mlnews"},{"@type":"ListItem","position":3,"name":"webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/webai-releases-twil-8b-an-8-8b-reasoning-model-scoring-98-2-on-math-500-ymqq9zelv#faq","mainEntity":[{"@type":"Question","name":"How does webAI's TwIL-8B compare to Qwen3.5-9B on math and code benchmarks?","acceptedAnswer":{"@type":"Answer","text":"TwIL-8B scores higher on math and code but lower on knowledge tasks. It posts 98.2 on MATH-500 versus 96.6 for Qwen3.5-9B, and 81.6 on LiveCodeBench (verbose variant) versus 65.6. On AIME 2026 Qwen3.5-9B leads 92.5 to 88.3. Across 13 shared benchmarks, TwIL-8B wins 7-6 overall. Comparing small reasoning models before picking one? Track benchmark releases like this on daily.dev."}},{"@type":"Question","name":"How much GPU memory or disk space does TwIL-8B need to run locally?","acceptedAnswer":{"@type":"Answer","text":"The Q4_K_M GGUF build is 4.98 GiB and runs on CPU or a modest-VRAM GPU via llama.cpp, suitable for both the balanced and verbose variants. Full bf16 weights total 16.38 GiB, and a Q8_0 build is 8.70 GiB for the verbose variant only. All reported benchmark scores used bf16 weights, so quantized runs may score slightly lower. Planning local LLM deployments? Daily.dev surfaces model release specs like these as they ship."}},{"@type":"Question","name":"Is TwIL-8B licensed for commercial use?","acceptedAnswer":{"@type":"Answer","text":"No, TwIL-8B's weights are open on Hugging Face but licensed for non-commercial use only. The model is an 8.79B parameter dense reasoning model post-trained from IBM's Granite 4.2 8B, with no hosted inference provider serving it yet. Vetting model licenses before adoption? daily.dev tracks open-weight release terms as they land."}}]} ```