$2 / month

Comment and reply across TML.

Start or manage membership

Most pulsed

Latest from other desks

tml.the modern leap
Tech / AI

OpenAI Cuts Astra Price Twenty Percent With GPT-6.2

Astra drops to $8 in and $40 out, cached input falls 60 percent, and both new models carry OpenAI's Critical cyber rating.

TML Tech DeskNov 22, 4:41 PM ET · 2 min read
Server racks in a data center. GPT-6.2's price fight runs on cached tokens and cost per task. Image: Pexels (Pexels License), via https://www.pexels.com/photo/1490877/.
Summary

The sticker matched. The meter moved.

Tap a passage. React to the part that matters.

OpenAI released GPT-6.2 Astra and Sol on November 17 with a 20 percent cut to the flagship's API price.

The new card is specific. Astra lists at $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for GPT-6 Astra. Cached input falls from $1.00 to $0.40, a 60 percent cut that matters most to agents resending long prompts. Sol holds its predecessor's card exactly, at $2 in, $0.10 cached and $10 out, so the cheaper tier's upgrade arrives with no change to the bill.

Where the models live is the quieter half of the launch. Both sit in ChatGPT's Work tab and in Codex for Plus, Pro, Enterprise and Edu users, and developers reach them in the API as gpt-6.2-astra and gpt-6.2-sol. The ordinary Chat tab doesn't change, so the release reaches paying work users first and the free tier not at all.

OpenAI sells Astra on cost per finished task rather than cost per token, and it says the model sets new highs on professional and scientific evaluations at under a third of its predecessor's cost per task. Artificial Analysis, an independent scoring firm, measured one benchmark test at about $4 for Astra against $26.54 for Anthropic's Claude Fable 5.2. That gap, not the sticker, is the argument the price cut is built on.

The system card carries the caveat. OpenAI rates both models Critical for cyber capability under its Preparedness Framework, the highest tier it assigns, and it shipped them with tighter guardrails on higher-risk security help. Some security research requests that earlier models answered will draw refusals, and OpenAI hasn't published retirement dates for the GPT-6 models the new pair replaces.

The timing is the frame. The launch lands a week after Anthropic's Claude 5.2 and Google's Gemini 4 releases, and Astra's list price now matches Fable 5.2's $8 and $40 card exactly. The fight has moved from the sticker to the meter: cached-input discounts, batch work at half price, and a fast tier at twice the standard rate.

The conversation

A little more perspective.

Reading is free. Verified members can comment and reply for $2 a month.

The conversation.

Loading comments…

Keep your curiosity going.

Some links here are affiliate links. If you buy through one, we might make a commission — it never changes what you pay. As an Amazon Associate I earn from qualifying purchases.