We’re excited to share a big upgrade. GPT-5.2 and GPT-5.2 Pro are now available inside the Pickaxe Builder, ready for all your tools.
These models bring a noticeable jump in reasoning, structure, accuracy, and multi-step planning. Tasks that once needed heavy prompting now come out cleaner, more organized, and more reliable.
Below are a few examples showing what this upgrade looks like in action.
GPT-5.1 Thinking vs GPT-5.2 Thinking
A simple prompt for a workforce planning model produces very different outcomes. GPT-5.2 builds a full structured spreadsheet with clear logic and clean formatting.
OpenAI’s latest benchmarks show strong gains across reasoning, science, math, and software engineering. These improvements translate directly into stronger Pickaxe tools.
That was fast! I don’t even have 5.2 in my paid ChatGPT account yet. Way to go Pickaxe team!
My prediction about token costs going up is proving to be true. 5.2 is more expensive than 5.1. I predict all the model providers will continue to nudge up pricing as they start to move toward profitability.
Keep this in mind when switching to 5.2. And remember the rising prices are with the model providers. While your pickaxe bill is going up, that additional money is going to OpenAI.
Right now, the best token price is Grok 4.1 Fast Reasoning.
Grok 4.1 Fast Reasoning is not quite as smart as ChatGPT 5.1 and 5.2. It is perhaps 90% as smart for 10% the price. Grok 4.1 Fast Reasoning can handle up to 2 million tokens of context though making it great for document analysis.
A counterpoint to the “tokens getting more expensive” observation is that LLMs are getting more done with each token. ChatGPT’s 5.2 tokens cost way more than ChatGPT 3.5’s tokens, but they are also way better.
So instead of thinking “how much do tokens cost” the question should be “how many dollars does it cost to accomplish my goal?”
This second question is better for making decisions. But a lot of tasks were impossible with AI a year or two ago so it is hard to find a trend. Things went from “impossible” to “expensive.” Perhaps, over time, they will go from “expensive” to “less expensive”
Exactly. Very few people think about it this way, and I’m glad you called it out.
For anyone still confused, think of it like taking a trip. You can take a slow, cheap bus with multiple stops and delays. Or you can take a faster train that costs more per ticket but gets you straight to your destination.
Tokens are the ticket price. The model is the vehicle. What matters is the total cost to reach the destination, not the price of each ticket.
@abhi Ok, I am confused on the pricing for GPT 5.2 Pro. Your team had it set at one $ dollar sign now it is 4 $$$$ dollar signs! This can’t be accurate? Attaching screenshot here and highlighting in red that it says for only 100 uses that it costs $63.09. Please confirm if this is correct since it is even more expensive by multiples than Claude Opus 4.5. If it is, which is out of control, can you at least send an email out to Pickaxe customers to let them know when a new model is exponentially more expensive? This will give us fair warning so we can price according. Shouldn’t it be $6.31?
Yes! As @thomasumstattd mentioned above… it’s a very expensive model.
The pricing indications we provide on the pricing page and in the builder are super useful! You can see the full explanation here, but the price stated there is simply the median price across all runs using that model on our platform. It gets updated daily as more people use the models. We do this to keep things apples to apples and show the real, recognized price per use. But in the days right after a model is launched, this stat can vary a lot!
The $$ signs are just indicators, this isn’t really an actual listed price. If you’re curious about the actual pricing, you can take a look at OpenAI’s pricing docs here.
To see the price of any particular generation you should click the message insights icon after any message:
GPT-5.1 Thinking vs GPT-5.2 Thinking
Benchmark improvements
