BWhat changed
Released a customizable multimodal model family with documented asynchronous reinforcement learning, controllable token effort, calibration rewards and complementary rubric/factual-claim verification training.
Scientific scope
Released a customizable multimodal model family with documented asynchronous reinforcement learning, controllable token effort, calibration rewards and complementary rubric/factual-claim verification training. Institution-authored technical account/model-card evidence, not a paper-level theorem or independently reproduced calibration claim. Inherits MoE, Muon, distillation and proper scoring rules. Full/Small variants bundled; model-card architecture details can differ from launch descriptions.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Institution-authored technical account/model-card evidence, not a paper-level theorem or independently reproduced calibration claim. Inherits MoE, Muon, distillation and proper scoring rules. Full/Small variants bundled; model-card architecture details can differ from launch descriptions.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Inkling: Our Open-Weights Model
- Thinking Machines Lab: Inkling calibrated multimodal training resource — official supporting account
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction