I’ve been trying to find the right balance of AI models to use in Micro.blog. Currently we use a few different flavors of GPT, including OpenAI’s cheaper “flex” version when speed doesn’t matter much, such as in the background work that powers Inkwell’s Reading Recap feature.

API pricing is a little all over the map, so I thought I’d capture some of the recent prices here, for OpenAI and Anthropic. I’m including a couple older models too, like the retired GPT-4.5. This is per million tokens, input / output.

GPT-5.5:      $5.00 / $30
GPT-5.4:      $2.50 / $15
GPT-5.4 Mini: $0.75 / $4.50
GPT-5.4 Nano: $0.20 / $1.25
GPT-5:        $1.25 / $10
GPT-4.5:     $75.00 / $150
Opus 4.7:     $5.00 / $25
Opus 4.6:     $5.00 / $25
Opus 4:      $15.00 / $75
Sonnet 4.6:   $3.00 / $15
Haiku 4.5:    $1.00 / $5

What can we learn from this? I have no idea. Clearly GPT-4.5 was a gigantic model, and so OpenAI changed course into more efficient models. Opus has generally been more expensive, but now is comparable to GPT-5.5.

When it was teased this morning, I didn’t expect GPT-5.5 to actually ship today. Sounds like it’s a very good, efficient model:

Across these domains, GPT‑5.5 is not just more intelligent; it is more efficient in how it works through problems, often reaching higher-quality outputs with fewer tokens and fewer retries. On Artificial Analysis’s Coding Index, GPT‑5.5 delivers state-of-the-art intelligence at half the cost of competitive frontier coding models.

And from Sam Altman, posting on the social network that just won’t go away despite how much we’d like it to:

We believe in democratization. We want people to be able to use lots of AI; we aim to have the most efficient models, the most efficient inference stack, and the most compute. We want our users to have access to the best technology and for everyone to have equal opportunity.

Sora also officially shuts down in a few days, freeing up those resources.