Thinking Machines Lab publishes 'A Safe Path to Open Weights'

The lab's post argues open-weight models are public goods, but release is irreversible; it details how Thinking Machines assessed its Inkling model and proposes staged access, testing, and stronger defenses to widen access safely.