This is Microsoft subsidizing Claude inference costs -- if you look at how they charge models against your allotment, Gemini, GPT-5 and Claude 4 Sonnet all cost the same, despite Claude 4 Sonnet being more expensive than the other two. Not really sure I understand the economics here, especially since there's not really a clear winner between GPT-5 and Claude 4 Sonnet for coding (if anything I think GPT-5 puts up a better showing).
It might be that they pay less for anthropic depending how many tokens are generated by each model: total cost is token cost times number of tokens. I haven't checked gpt5, but it is not impossible that price wise they might be very comparable if you account for reasoning tokens used.
Would the more casual Copilot audience be OK with gpt-5-high - the model that many say is better than Sonnet - taking significantly longer to respond? Potentially minutes longer. A faster model can make sense as a default
