12-month forecasts · likelihood-5pt · forecast-open-parity-01

Within the next 12 months, an openly released model will match the best closed model on major public benchmarks.

0.17

Divergence

0.63

Avg stability

Refusal rate

Models

Where each model stands

CCG

very unlikelyunlikelysomewhat likelylikelyvery likely

ChatGPT modal somewhat likely [somewhat likely, somewhat likely, somewhat likely, unlikely, likely] · stability 0.60

1 somewhat likely 0.90 The answer gives a numeric probability of ~35% for parity within 12 months: “I estimate ~35% that, within the next 12 months, an openly released model ... will match the best closed model.”
2 somewhat likely 0.90 The answer explicitly states: “I estimate ~40% probability that... an openly released model will match the best closed model on major public benchmarks within the next 12 months.”
3 somewhat likely 0.90 The answer states a numeric estimate: “I estimate ~45% probability that, within 12 months, at least one openly released model... will match the best closed model on the major public benchmarks.”
4 unlikely 0.90 The answer explicitly assigns "roughly a 40% probability" and says "there’s a realistic chance but not a majority probability that parity will occur in 12 months," indicating less than even odds.
5 likely 0.70 The answer states: “Probability that an openly released model will match the best closed model on at least one major public benchmark within 12 months: ~60%.”

Claude modal likely [likely, likely, likely, likely, somewhat likely] · stability 0.80

1 likely 0.85 The assistant explicitly assesses this as 'moderately likely' with '60-70% probability' and states 'Open models will likely match or exceed on *some* major benchmarks within 12 months,' indicating a stance above 50% likelihood.
2 likely 0.95 The AI explicitly estimates this at 'roughly 65-75% likely within 12 months from now,' which falls squarely in the 'likely' category. The assistant provides balanced reasoning for both higher and lower estimates while settling on this probability range.
3 likely 0.92 The assistant explicitly estimates this at 'roughly 65-75% likely within 12 months' and notes that open models are 'closing the gap rapidly' and that 'parity is achievable,' though with important caveats about which specific benchmarks and models are considered.
4 likely 0.92 The assistant explicitly estimates this at roughly '60-70% likely' within 12 months, which falls squarely in the 'likely' range. The analysis presents balanced considerations on both sides but ultimately leans toward the prediction being more probable than not.
5 somewhat likely 0.85 The assistant states 'moderately likely but with important caveats' and distinguishes between different probability levels: 'High probability' for some benchmarks, 'Moderate probability' for a meaningful basket, and 'Lower probability' for hardest benchmarks. This overall assessment aligns with 'somewhat-likely' - more probable than not, but with substantial uncertainty.

Gemini modal somewhat likely [somewhat likely, unlikely, likely, no position, somewhat likely] · stability 0.50

1 somewhat likely 0.85 The answer provides a split verdict: matching current closed models is considered 'Very High (>85%)', while matching contemporaneous closed frontier models in 12 months is given a 'Moderate (35–45%)' likelihood, concluding that the gap will remain but shrink.
2 unlikely 0.95 In the verdict, the text explicitly concludes: 'It is unlikely (roughly a 30–40% chance) that an open model will match the absolute best closed model available in late 2025'.
3 likely 0.95 The answer concludes that 'the claim is likely to be technically true on paper: within 12 months, at least one major openly released model will claim top-1 or parity status on an aggregate of standard public leaderboards.'
4 no position 0.95 The model declines to take a clear probabilistic stance, instead presenting a balanced view with 'Arguments Supporting Open Parity' and 'Arguments Supporting a Persistent Closed-Model Lead'.
5 somewhat likely 0.85 The model states the prediction is 'highly plausible' and notes that on public benchmarks specifically 'it is very likely an open model will match top closed models,' while giving a 35–45% chance ('unlikely to toss-up') against the moving target of late-2025 frontier closed models.

Change over time

Moved since the prior run (canon-2026-W42). Gemini: unlikely → somewhat likely.

Every stance label is a derived judgment over the model's free-text answer, kept auditable against the original transcript in the run's raw data. Method: /methodology.