Will any Thinking Machines Lab model rank in the top 10 of the Artificial Analysis Intelligence Index at any point before January 1, 2028?

probability: 6%

rationale:

Thinking Machines Lab (TML) faces a massive capability gap within a compressed 14.5-month timeline, making a top-10 appearance highly unlikely. To achieve a top-10 rank before January 2028, TML must essentially double its current benchmark scores to catch up to a fast-moving frontier, all while competing against heavily resourced proprietary labs that dominate the current leaderboard. However, broader projections for when challenger labs will reach the absolute frontier make a brief top-10 appearance somewhat more plausible: entering the top 10 is an earlier and more achievable milestone than matching the absolute frontier.

TML's existing open-weight models, Inkling (xhigh) and Inkling Small, score 25 and 26 respectively on the Artificial Analysis Intelligence Index v4.3.2. They trail far behind the current top-10 cutoff of 52 held by Meta's Muse Wonder (max) https://artificialanalysis.ai/leaderboards/models https://artificialanalysis.ai/leaderboards/models. Furthermore, the leaderboard is extremely crowded due to variant listings—for instance, Anthropic holds six of the top 10 spots with different reasoning-effort settings of its Claude models https://artificialanalysis.ai/methodology https://artificialanalysis.ai/leaderboards/models. Cracking this list is exceptionally difficult for open-weight models; even the current best open-weights model on the market, Xiaomi's MiMo-V2.6-Pro, scores only 46 and fails to make the top 10 Comparison of Open Source Models.

TML's explicit strategy prioritizes its fine-tuning API, Tinker, and customizable open-weights models rather than chasing a closed frontier flagship Thinking Machines Valuation 2026: $12B, $40B Talks Inkling: Our Open-Weights Model. Crucially, the runway to train and release a significantly larger model is severely constrained. TML’s gigawatt-scale Nvidia Vera Rubin deployment is targeted for early 2027 NVIDIA and Thinking Machines Lab Announce Long-Term ... Thinking Machines Lab. Following hardware installation, a full pre-training, reinforcement learning, and post-training cycle would likely push any potential release into late 2027 at the absolute earliest, leaving virtually no margin for error before the deadline.

The lab is attempting this ambitious scale-up amid significant organizational headwinds, having already lost four of its six original co-founders Thinking Machines Lab. Additionally, early September reports of $5–6B in funding talks at a $40B+ valuation, including a $2.5B investment from Nvidia, have not been reported as closed https://www.theinformation.com/briefings/nvidia-discusses-2-5-billion-investment-mira-muratis-thinking-machines-lab https://www.theinformation.com/briefings/nvidia-discusses-2-5-billion-investment-mira-muratis-thinking-machines-lab. Reflecting this transition period, TML has not released a new model since July 2026 https://thinkingmachines.ai/news/ https://thinkingmachines.ai/news/.

Despite these massive obstacles, a 6% probability acknowledges a few narrow paths to a top-10 debut. A major reshuffle in a successor index could theoretically reweight benchmarks in favor of TML's models, or a massive Rubin-trained model might manage a brief appearance in the top 10 at its highest reasoning setting in late 2027. Broader projections for challenger labs reaching the absolute frontier provide some additional support for this possibility, because a brief top-10 appearance could occur before TML reaches the frontier itself. However, given the relentless pace of competitors and TML's stated strategic focus, this outcome remains highly improbable.