ChatGPT Fast & cheap Verified 19 Aug 2026

GPT-5.6 Luna

The cheap, fast tier for volume work — 1/25th of Sol's input price.

Maker: OpenAI
Context: 1.05M tokens
Pricing: $0.20 / 1M tokens / $1.20 / 1M tokens

Cut from $1 / $6 on 30 Jul 2026, an 80% drop. Long-context requests are re-rated to $0.40 / $1.80. The cheapest way to touch a 1M-token window on any major provider, by a wide margin.

Best for

Classifying, tagging, or extracting at volume
High-throughput automations where cost-per-call compounds
Long-document scanning where you only need retrieval, not judgement

Prompting quirks

Give it a rubric, not a judgement call

Budget tiers reward unambiguous targets. Define the categories, the output format, and an example; don't ask it to decide what matters.

The cheap tier is where big context actually pays

A 1M window at $0.20 per 1M input tokens changes what's worth automating. Scan-and-extract jobs you'd never run on a flagship become routine here.

Copy-paste example

Tuned for GPT-5.6 Luna. Swap in your own content where marked.

Extract commitments from a meeting transcript

From the transcript below, list every commitment someone made.
Output as "owner | commitment | date mentioned (or 'none')".
No commentary, no summary. If there are none, say "none".
<transcript>
{paste transcript}
</transcript>

Facts verified against the official OpenAI docs on 19 August 2026.