The performance of the GPT-5 model varies greatly based on the reasoning-effort parameter. With minimum, it is a very unreliable, low rank model. With high, it is the #1 on many benchmarks.
Is it possible to make sure / assert that OpenCode is using GPT-5 with high reasoning effort?
The performance of the GPT-5 model varies greatly based on the reasoning-effort parameter. With minimum, it is a very unreliable, low rank model. With high, it is the #1 on many benchmarks.
Is it possible to make sure / assert that OpenCode is using GPT-5 with high reasoning effort?