{
  "site": "Agentic Weekly",
  "dataset": "Agentic Weekly model facts",
  "as_of": "2026-11-05",
  "license": {
    "name": "CC BY 4.0",
    "url": "https://creativecommons.org/licenses/by/4.0/"
  },
  "license_spdx": "CC-BY-4.0",
  "attribution": "Data: Agentic Weekly (https://theagenticweekly.com), CC BY 4.0",
  "usage": "Quoting, summarizing and citing this data is encouraged. Credit Agentic Weekly (https://theagenticweekly.com), link the page you used and keep its Reported or Tested label.",
  "page": "https://theagenticweekly.com/models",
  "evidence_labels": {
    "Reported": "Read from the cited vendor page on the as-of date; not run by us.",
    "Tested": "We ran it ourselves; scope is stated in tested_note. No model is Tested yet."
  },
  "provenance": {
    "note": "Every fact object carries its own provenance: fact, value, source_url, source_type, verified_at, evidence_status and, for tested facts, test_id. Read it per fact, not per page. Full explanation: https://theagenticweekly.com/methodology#provenance",
    "source_type": {
      "vendor_reported": "Read from the vendor's own page (docs, pricing, release notes, post); not run by us.",
      "independently_tested": "We ran it; test_id names the test log. Smoke tests are not benchmarks.",
      "editorial": "Our own judgment or a ranking built on stated criteria; never presented as a vendor fact."
    },
    "evidence_status": {
      "confirmed": "The cited source_url states this value; we read it on verified_at.",
      "not_confirmed_on_source": "The cited page does not itself state this value; the note says what we infer and from where.",
      "promo": "The price is temporary; pricing.list holds the standing price and pricing.promo_label says when the promo ends.",
      "partial": "Part of the value was read on the cited page.",
      "not_applicable": "Editorial judgment or a relationship disclosure, not a checkable fact."
    },
    "verified_at": "The date we last re-read the cited page. Only models re-read on a date carry that date; others keep their earlier date."
  },
  "model": {
    "slug": "twil-8b",
    "name": "TwIL-8B",
    "vendor": "webAI",
    "family": "TwIL",
    "category": "llm",
    "status": "generally-available",
    "groups": [
      "open-weights"
    ],
    "short_answer": "TwIL-8B is webAI's latest language model in our dataset, released Nov 4, 2026. Open weights (webAI Non-Commercial License ver. 1.0). webAI's 8.8B reasoning model for math, code and instruction following, released Nov 4, 2026, post-trained from IBM Granite 4.2 8B under a non-commercial license. Verified against vendor sources on Nov 4, 2026.",
    "summary": "webAI's 8.8B reasoning model for math, code and instruction following, released Nov 4, 2026, post-trained from IBM Granite 4.2 8B under a non-commercial license.",
    "release_date": "2026-11-04",
    "latest_update": null,
    "api_id": null,
    "context_tokens": 131072,
    "max_output_tokens": null,
    "input_modalities": [
      "text"
    ],
    "output_modalities": null,
    "parameters": {
      "total": "8.79B",
      "architecture": "dense decoder-only Transformer (Granite)"
    },
    "license": "webAI Non-Commercial License ver. 1.0",
    "open_weights": true,
    "pricing": null,
    "evidence": "Reported",
    "tested_note": null,
    "last_verified": "2026-11-04",
    "is_stale": false,
    "facts": {
      "release_date": {
        "fact": "release_date",
        "value": "2026-11-04",
        "source_url": "https://huggingface.co/webAI-Official/TwIL-8B",
        "source_type": "vendor_reported",
        "verified_at": "2026-11-04",
        "evidence_status": "confirmed",
        "as_of": "2026-11-04",
        "evidence": "Reported",
        "note": "Date of the public model card update and launch coverage; the card has no explicit release timestamp.",
        "anchor_url": "https://theagenticweekly.com/models/twil-8b#fact-release-date"
      },
      "context_tokens": {
        "fact": "context_tokens",
        "value": 131072,
        "source_url": "https://huggingface.co/webAI-Official/TwIL-8B",
        "source_type": "vendor_reported",
        "verified_at": "2026-11-04",
        "evidence_status": "confirmed",
        "as_of": "2026-11-04",
        "evidence": "Reported",
        "note": "Inherited from the Granite 4.2 8B base model, per the model card.",
        "anchor_url": "https://theagenticweekly.com/models/twil-8b#fact-context-tokens"
      },
      "input_modalities": {
        "fact": "input_modalities",
        "value": [
          "text"
        ],
        "source_url": "https://huggingface.co/webAI-Official/TwIL-8B",
        "source_type": "vendor_reported",
        "verified_at": "2026-11-04",
        "evidence_status": "confirmed",
        "as_of": "2026-11-04",
        "evidence": "Reported",
        "anchor_url": "https://theagenticweekly.com/models/twil-8b#fact-input-modalities"
      },
      "parameters": {
        "fact": "parameters",
        "value": {
          "total": "8.79B",
          "architecture": "dense decoder-only Transformer (Granite)"
        },
        "source_url": "https://huggingface.co/webAI-Official/TwIL-8B",
        "source_type": "vendor_reported",
        "verified_at": "2026-11-04",
        "evidence_status": "confirmed",
        "as_of": "2026-11-04",
        "evidence": "Reported",
        "anchor_url": "https://theagenticweekly.com/models/twil-8b#fact-parameters"
      },
      "license": {
        "fact": "license",
        "value": {
          "type": "open-weights",
          "name": "webAI Non-Commercial License ver. 1.0",
          "openWeights": true
        },
        "source_url": "https://huggingface.co/webAI-Official/TwIL-8B",
        "source_type": "vendor_reported",
        "verified_at": "2026-11-04",
        "evidence_status": "confirmed",
        "as_of": "2026-11-04",
        "evidence": "Reported",
        "anchor_url": "https://theagenticweekly.com/models/twil-8b#fact-license"
      }
    },
    "vendor_claims": [
      "webAI reports 98.2 on MATH-500 and 81.6 on LiveCodeBench (verbose variant) for the 8.79B model. Vendor claim; we did not run it."
    ],
    "notes": [
      "Ships as two variants, balanced (default) and verbose, plus GGUF builds from 4.98 GiB (Q4_K_M) to 16.38 GiB (F16)."
    ],
    "not_verified": [
      "No hosted inference provider or price is listed; the model is weights-only on Hugging Face.",
      "Benchmark scores are webAI's own internal runs, and several were measured only on the verbose variant, so they are listed as vendor claims only."
    ],
    "sources": [
      {
        "label": "Hugging Face model card",
        "url": "https://huggingface.co/webAI-Official/TwIL-8B"
      },
      {
        "label": "Agentic Weekly brief",
        "url": "https://theagenticweekly.com/news/twil-8b-open-weight-reasoning-model"
      }
    ],
    "url": "https://theagenticweekly.com/models/twil-8b",
    "primary_source_url": "https://huggingface.co/webAI-Official/TwIL-8B",
    "anchors": {
      "facts": "https://theagenticweekly.com/models/twil-8b#facts",
      "sources": "https://theagenticweekly.com/models/twil-8b#sources",
      "tested": "https://theagenticweekly.com/models/twil-8b#tested",
      "not_verified": "https://theagenticweekly.com/models/twil-8b#left-out"
    },
    "markdown": "https://theagenticweekly.com/models/twil-8b.md",
    "api": "https://theagenticweekly.com/api/models/twil-8b"
  }
}