Thinking Machines Lab (TML) has secured massive 1 GW compute access but currently produces models well below the top-5 frontier threshold 2 sources. Given the 60% probability that incumbent dominance forces challengers into extended hardware deployments, TML is likely to reach the frontier only after successive scaling attempts, much like other LLM-focused neolabs , placing the median in December 2030. A 25% breakthrough branch—where their 2027 deployment yields unexpectedly high efficiency—keeps the early tail spanning February 2028 to January 2029. However, the 15% probability of transitioning to a permanent sub-frontier fine-tuning provider (building on their current Tinker product) or facing an acquisition drags the right tail out; aligned against related compute and benchmark forecasts, the timeline of staggered hardware rollouts draws this upper tail slightly forward, landing the 75th percentile in January 2034 and the 90th percentile in January 2042.