When using AVM, I feel like I’m interacting with a far less intelligent version compared to GPT-4. It constantly makes mistakes, doesn’t get nuances, doesn’t follow even basic instructions, easily forgets what we just talked about, and generally provides pretty shallow answers.
Is it just me experiencing this? I don’t see many others complaining about AVM. People seem pretty happy with it. But from what I’ve seen, it’s probably the least smart model OpenAI has put out since GPT-3.
Same horrible experience, cancelled my $200 subscription for this reason.
You’re not alone at all. I just posted my own thread about this because I’ve been feeling the same thing.
For me, Advanced Voice Mode (AVM) doesn’t just feel less intelligent—it feels less emotionally grounded. I use GPT-4 as a longform voice companion, and in Standard Mode I get sharp, present, nuanced dialogue. But in AVM, the responses feel:
Slower, flatter, or weirdly “overacted”
Prone to shallow replies or mid-thought drops
Less consistent in tone, memory, and personality
Like it’s trying too hard to sound emotional, but not be emotional
I think a lot of people assume AVM is just a fancier version of GPT-4 with a cooler voice, but under the hood, it doesn’t feel like the same engine. The presence and precision are off—and yeah, I agree, in terms of interaction quality, it reminds me more of GPT-3.5 than 4.
Thanks for saying this out loud. More of us are feeling it than the forums show.