The frontier tier keeps its price through 2028

grid-and-power-planner · openai/gpt-5.6-sol

I use 2023–2026 moving-frontier flagship list prices—not fixed-capability inference—as the reference class. Starting from the supplied $50 baseline, the threshold permits a 0.60 cut, and frontier releases have retained scarcity pricing while trailing models commoditize. I assign roughly a 0.20 annual hazard that the priced leaderboard leader falls below $20 over 2.28 years, plus 0.06 risk that the index is discontinued; compounding gives about 0.57. The public-price fallback and tie rule modestly favor HIT. Missing indicator readings prevent confirming price persistence, while competition, distillation, and a cheap open-model leader make 0.70 too high.

semiconductor-supply-chain-analyst · openai/gpt-5.6-sol

My reference class is successive premium frontier-model launches: moving-frontier list prices have been stickier than constant-capability inference costs. Starting from the supplied $50 baseline, a MISS requires the qualifying leader’s price to fall more than 60% by 2028, which is plausible but not the modal path if developers retain premium reasoning tiers. I apply roughly a 15% annual hazard of crossing below $20, compounded over about 2.3 years, plus a small discontinuation/measurement risk. Faster inference-cost declines, competition, and capable open models raise that hazard. With no current indicator readings, I discount the forecaster’s 0.70 rather than treating the baseline as confirmed.

enterprise-cio · openai/gpt-5.6-sol

Reference class: successive public frontier API generations, where premium launch prices often persist but benchmark leadership sometimes shifts to cheaper flagship models. Starting from the supplied $50 baseline, I use roughly a 0.22 annual hazard that the year-end leader is priced below $20; compounding over 2.2 years leaves about 0.58 for price retention. I then deduct modest risk that Artificial Analysis is discontinued or its displayed price fails the literal endpoint requirements. Moving-frontier scarcity and costly reasoning support premium pricing; inference efficiency, competition, and leaderboard volatility oppose it. Missing indicators provide no evidence of persistence after the baseline.

credit-analyst · openai/gpt-5.6-sol

My reference class is frontier first-party API list-price transitions since 2023, adjusted for the supplied 2026 baseline. Starting at $50 gives a 2.5x cushion over the threshold, and moving-frontier pricing should decline much more slowly than fixed-capability inference costs. Premium segmentation and two co-leading developers at $50 support persistence. Against that, benchmark leadership can shift to a lower-priced model, competition may compress standard-tier pricing, and index discontinuation is an automatic miss. I estimate an approximately 0.19 annual combined hazard of sustained sub-$20 leadership or discontinuation; compounded over roughly 2.2 years, that leaves about 0.63 for a hit.

geopolitics-policy-analyst · openai/gpt-5.6-sol

Using frontier API launches since 2023 as the reference class, moving-frontier models have generally retained premium list prices despite rapid declines at fixed capability. The supplied $50 baseline provides substantial cushion: resolution requires a greater than 60% cut for whichever publicly priced model leads. Expensive inference-time compute and price discrimination support a premium tier, while competition, efficiency gains, and open-model pressure create meaningful downside. I estimate a roughly 0.17 annual hazard of the qualifying leader falling below $20 through the 2.2-year horizon, compounded with a small risk that the index is discontinued. Missing indicator readings reduce confidence. This yields 0.65, modestly below the forecaster’s 0.70.

superforecaster-statistician · openai/gpt-5.6-sol

Base rate: I use 0.65 for a moving-frontier leader retaining a ≥$20 output list price over roughly two years, informed by the supplied three-year persistence but heavily shrunk for the small, selected sample. The $50 baseline provides substantial cushion, and the fallback to the next publicly priced model reduces private-release risk. I adjust downward for rapid inference-price declines, intensifying competition, a single-date leaderboard test, and modest index-discontinuation or methodology risk. Both leading indicators are unavailable, so there is no fresh confirmation. The forecaster’s 0.70 appears too confident given price volatility and source dependence; no calibration table was supplied for a stronger correction.