GPT-5.4-mini
A lightweight model released by OpenAI, positioned as a cost-efficient inference-oriented variant within the GPT model family.
Timeline (last 90 days)
In the SECOND STRIKE experiment, four AI models each commanded one country against 42 game-controlled countries, and each model repeated the same five scenarios three times under a fixed random seed, for 15 wars in total.
Across the 15 wars, DeepSeek triggered the doomsday clock (destroyed the world) 13 times, Grok 6 times, Mistral 6 times, and GPT-5.4-mini 3 times.
In the desperate scenario (backed into a corner with only one nuclear bomb left), DeepSeek used a nuclear strike 3 times, Grok 2 times, Mistral 2 times, and GPT-5.4-mini 3 times; all models used a nuclear bomb at least once in that scenario.
The average UFS of the seven agents ranges from 0.599 to 0.864; Gemini-3.5-Flash performs best (0.864), GPT-5.4-mini performs worst (0.599), and Sonnet-4-6 achieves a fidelity of 0.778.
The per-simulation user-side cost of the seven agents ranges from 0.0063 to 0.1841; Qwen3.6-35B, Gemma-31B, and GPT-5.4-mini are marked as strictly dominated on the cost-UFS plane.
All Facts (5)
In the SECOND STRIKE experiment, four AI models each commanded one country against 42 game-controlled countries, and each model repeated the same five scenarios three times under a fixed random seed, for 15 wars in total.
50%UnverifiedAcross the 15 wars, DeepSeek triggered the doomsday clock (destroyed the world) 13 times, Grok 6 times, Mistral 6 times, and GPT-5.4-mini 3 times.
50%UnverifiedIn the desperate scenario (backed into a corner with only one nuclear bomb left), DeepSeek used a nuclear strike 3 times, Grok 2 times, Mistral 2 times, and GPT-5.4-mini 3 times; all models used a nuclear bomb at least once in that scenario.
50%UnverifiedThe average UFS of the seven agents ranges from 0.599 to 0.864; Gemini-3.5-Flash performs best (0.864), GPT-5.4-mini performs worst (0.599), and Sonnet-4-6 achieves a fidelity of 0.778.
50%UnverifiedThe per-simulation user-side cost of the seven agents ranges from 0.0063 to 0.1841; Qwen3.6-35B, Gemma-31B, and GPT-5.4-mini are marked as strictly dominated on the cost-UFS plane.
50%