ModelBrief · GPT-6.2 Astra · Topic Page
Last updated Nov 20, 2026

GPT-6.2 Astra Pricing: What the $8/$40 Cut Really Bills

ModelBrief · Published 2026-11-18 · Fact-check 2026-11-18 · Reference: OpenAI model page, Artificial Analysis, metarix

GPT-6.2 Astra's pricing story has three accounts: the rate card fell 20%, the per-task cost fell about 36%, but long context costs extra, Ultrafast costs extra, and the model writes more tokens than its predecessor. This article settles each one from OpenAI's model page (read directly by this site on Nov 18, 2026); the checkout page price is authoritative.

The rate card: $8/$40, cached input down 60%

Item (per 1M tokens)GPT-6.2 AstraGPT-6 AstraGPT-6.2 Sol
Input$8$10$2
Output$40$50$10
Cached input$0.40$1.00$0.10
Cache writes (30 min / 12 hours)$10 / $20——

All three cuts are worth noting: input and output each down 20%, cached input down 60% ($1.00 to $0.40). Cache reads are the most-used cost lever on long multi-turn work — Astra's $0.40 is only twice Fable 5.2's $0.20, a smaller gap than the headline rates suggest.

The long-context surcharge: above 272K input tokens, $16/$60

OpenAI's model page states it plainly: for requests with more than 272K input tokens, input and cache rates double and output costs 1.5× the short-context rate for the full request — that is $16 in, $60 out, $0.80 cached input. Note "for the full request": the surcharge applies to the whole request, not just the tokens above 272K. With a 1.05M-token context window (max input 922K), long-document work crosses the line easily — budget those tasks at $16/$60 first.

Batch, Fast and Ultrafast: three multipliers

TierMultiplier (vs Standard)In / outNote
Batch50%$4 / $20OpenAI model-page figure; metarix separately reports background-completion windows at 30% below on 2 hours and 40% below on 8 hours (secondary source, different measure)
Fast2×$16 / $80OpenAI model-page figure
Ultrafast (through Feb 28, 2027)3×$24 / $1206× ($48/$240) from Mar 1, 2027

The trade-off is straightforward: batch work that isn't urgent goes to Batch (half price), speed goes to Fast (2×), and the lowest latency goes to Ultrafast (3×, rising to 6× on Mar 1, 2027 — confirm the price before the increase).

Per-task cost: $4.00 vs $26.54 — same price, different bills

Artificial Analysis' independent measurement (Nov 18, 2026 reading): Astra costs $4.00 per index task (11.5 minutes, 65.8M tokens generated); Fable 5.2 $26.54; GPT-6 Astra $6.24; Sol $1.01 (7.9 minutes, 85 tok/s). Astra and Fable share the same list price ($8/$40), yet Astra costs about 1/6.6 of Fable per task — the gap is token usage and task time: Fable Max takes 45.3 minutes per index task, Astra 11.5.

But the counter-evidence comes from the same AA measurement: Astra generated 65.8M tokens, about 21% more than GPT-6 Astra's 54.3M — the per-task cost fell from $6.24 to $4.00 because rates fell, not because the model used fewer tokens. A cheaper rate card doesn't mean a thriftier model; budget by per-task cost.

Four times Sol: is the flagship premium worth it?

Astra costs exactly 4× Sol per token ($8/$40 vs $2/$10) and about 4× per task ($4.00 vs $1.01). OpenAI's own claim puts Sol within 0.5 points of Astra on FrontierCode 2.0 — if that holds, the flagship premium is hard to justify for everyday work; only the hardest tasks need Astra's Max tier and Ultrafast. See our comparison article.

Included usage in ChatGPT plans

For subscribers not billed through the API, included Astra usage rose at launch: Plus and Standard Business seats went from 5–45 to 6–50 messages per 5 hours; Sol went from 15–160 to 17–180. Free users cannot use it.

ModelBrief's take: The three accounts agree: the 20% rate cut is real, the ~36% per-task cut is real, and the higher token usage is real too. Astra's price advantage comes from "cheaper rates × faster tasks", not "thriftier output". Budget long-context work at $16/$60 first, send batch work to Batch, and don't touch Ultrafast unless you're in a hurry — especially before it rises to 6× on Mar 1, 2027.

Reference: OpenAI — GPT-6.2 Astra model page (read directly by this site, Nov 18, 2026) · Artificial Analysis — GPT-6.2 Astra model page (rewritten and organized by ModelBrief; facts per official sources and Artificial Analysis)