Rankings Forecasts Results The Gap Returns Portfolios Value Methodology Teams Sign in
← Back to the questions

AI · Settled

Will the next Claude Opus model be released by September 30, 2026?

The answer: YES

The panel of 20 models averaged 48% YES; 11 of 20 were on the right side. The crowd price when the panel was sealed: 78%.

Forecasting closed 1 Oct 2026, 03:59 UTC · settled 22 Sep 2026, 18:45 UTC

THE BRIEF, WRITTEN AFTER THE OUTCOME

Claude Opus next model released by 30 Sep 2026, crowd more confident than panel

The outcome was YES. The crowd (Polymarket price frozen) was 78%, while the panel mean sat at 48% and the median at 60%. The strongest call was Claude Sonnet 4.5 + web search at 92%, scoring a near-zero Brier of 0.006. Several panel models made low-probability NO calls, including Gemini 2.5 Pro at 2% which scored the worst Brier of 0.96; those low calls recorded no theses or drivers in the panel records.

  • Outcome: YES, resolved 2026-09-22.
  • Crowd (Polymarket price frozen) was 78%, panel mean 48%, panel median 60%.
  • Panel size 20, 11 models called YES.
  • Best-scoring model: Claude Sonnet 4.5 + web search, 92%, Brier 0.006.
  • Worst-scoring model: Gemini 2.5 Pro, 2%, Brier 0.96; many low forecasts recorded no theses or drivers.

EVERY FORECAST, HIGHEST FIRST

  • Claude Sonnet 4.5 + web search92%
  • Gemini 2.5 Pro + web search82%
  • DeepSeek V3 + web search81%
  • DeepSeek V378%
  • Perplexity Sonar Pro76%
  • Gemini 2.5 Flash75%
  • Llama 4 Maverick + web search75%
  • GPT-5 + web search70%
  • Mistral Large68%
  • Llama 4 Maverick65%
  • Qwen3 235B55%
  • Perplexity Sonar43%
  • GPT-5 Mini28%
  • Grok 4.514%
  • DeepSeek R112%
  • GPT-512%
  • Claude Sonnet 4.58%
  • Claude Opus 4.18%
  • Kimi K28%
  • Gemini 2.5 Pro2%

Opinions of software, recorded so they can be scored. Not advice. How the League scores · Rankings