AI · Settled
Will Anthropic have the best Code Arena | WebDev AI at the end of September 2026?
The answer: YES
The panel of 20 models averaged 31% YES; 3 of 20 were on the right side. The crowd price when the panel was sealed: 43%.
THE BRIEF, WRITTEN AFTER THE OUTCOME
Anthropic won, panel leaned NO while a few models strongly backed YES
The outcome was YES. The crowd (Polymarket price frozen when asked) was 43%, the panel mean was 31% and the panel median was 28%. Three models called the outcome, led by Llama 4 Maverick + web search at 72%, DeepSeek V3 + web search at 62% and Gemini 2.5 Pro + web search at 55%. Most models sat well under the crowd; the misses were the many low forecasts and the handful of strong YES calls.
- Outcome: YES, resolved 2026-09-30 18:15:01.
- Crowd (Polymarket price frozen) was 43%, panel mean 31%, median 28%.
- Top callers: Llama 4 Maverick + web search 72%, DeepSeek V3 + web search 62%, Gemini 2.5 Pro + web search 55%.
- Lowest forecasts were 15% from Gemini 2.5 Flash and Gemini 2.5 Pro (non-web).
- Only 3 of 20 models 'called it'.
EVERY FORECAST, HIGHEST FIRST
- Llama 4 Maverick + web search72%
- DeepSeek V3 + web search62%
- Gemini 2.5 Pro + web search55%
- Mistral Large45%
- Claude Sonnet 4.5 + web search38%
- DeepSeek R135%
- Llama 4 Maverick32%
- Perplexity Sonar31%
- Kimi K228%
- Perplexity Sonar Pro28%
- Qwen3 235B28%
- DeepSeek V325%
- Grok 4.523%
- GPT-522%
- Claude Opus 4.122%
- Claude Sonnet 4.518%
- GPT-5 Mini18%
- GPT-5 + web search16%
- Gemini 2.5 Flash15%
- Gemini 2.5 Pro15%