OpenAI measures its 50% GPT-6 price cut against the previous generation's promotional rate
Sol lists at $2 and $10 per million tokens and Luna at $0.10 and $0.50. The per-task savings OpenAI published come mostly from the lower price, and the cheaper tier scores below its predecessor on computer use.
The Scientist · Science desk
What happened
- OpenAI added two cheaper tiers to the GPT-6 family, Sol and Luna, and said it cut their API prices by half compared with what it describes as GPT-5.6 promotional pricing.
- Sol lists at $2 per million input tokens and $10 per million output and Luna at $0.10 and $0.50, one fifth and one hundredth of Astra's $10 and $50 respectively.
- xAI priced Grok 4.7 at $2 and $6 on September 21, Anthropic released Opus 5.5 at $4 and $20 the next day, and OpenAI published Sol and Luna ninety minutes after that.
Why it matters
- What it costs cost Anyone who budgeted this quarter on GPT-5.6 promotional rates gets half the per-token price and an unquantified change in tokens per task, so the saving that shows up on the invoice is not 50 percent.
- What it limits constraint A team running desktop agents on GPT-5.6 Sol cannot treat the cheaper tier as a drop-in, because the same workload gives up 5.2 points of measured completion for the lower price.
- Who has to decide decision OpenAI's own reported drop to 1.3 percent coding deception in a tier priced at a fifth of Astra makes it harder for any vendor to argue that frontier-tier prices are what buys behavioural safety.
Next story · Product · Sep 30
OpenAI's 10-cent GPT-6.1 Sol price covers only cached inputConfirmed5 publishers
Per-token price and cost per task are different numbers, and the second one lands on the invoice. OpenAI's own figures show the gap. On AutomationBench, GPT-6 Luna at high effort scored 5.4 percentage points above its predecessor at 58 percent lower cost per task [16]. The list price fell by half, against GPT-5.6 promotional pricing in OpenAI's wording, and the post does not give the standard rates [2]. Cost per task is price times tokens consumed. If both published figures hold and the mix of input and output tokens is unchanged, Luna spent about 16 percent fewer tokens per task than GPT-5.6 Luna did: 0.42 divided by 0.50 is 0.84 [26].
The effort setting accounts for the rest. Every comparison OpenAI leads with is run at xhigh or max effort [8][17]. Forkast puts Sol at 33.2 percent on AutomationBench at $0.27 per task, against Claude Opus 5's 26.9 percent at roughly $3 [6]. That matches OpenAI's claim of 9 percent of Opus 5's cost per task [7]. At Sol's $10 per million output tokens, $0.27 covers at most 27,000 output tokens, so one "task" on that test is a long run [27]. OpenAI also reports Luna at max effort scoring 66.6 percent on DeepSWE v1.1, 93 percent cheaper per task than Opus 5 and 96 percent cheaper than Fable 5 [9].
On output tokens, rivals undercut OpenAI. Sol's $2 input price matches Grok 4.7 [13], but Grok's output price is $6 and Sol's is $10, so Sol costs 67 percent more per output token [24]. Luna's $0.10 input undercuts Xiaomi's MiMo-V2.6 Flash at $0.14, while Luna's $0.50 output is 79 percent above MiMo's $0.28 [29]. Forkast writes that Luna "sits below every frontier-class model on the market" [28]. For output-heavy work, its own price list says otherwise.
On computer use the cheaper model did worse. Forkast reports Sol at xhigh scoring 60.5 percent on OSWorld 2.0 against GPT-5.6 Sol's 65.7 percent [10], a fall of 5.2 points [19]. OpenAI says Astra remains its best model for computer use and that Sol and Luna "offer more cost-efficient performance than their predecessors" [18]. Both can be true at once, and a team pointing an agent at a desktop is trading measured completion for the lower price.
OpenAI's internal factuality evaluation is built from de-identified real conversations in which users flagged mistakes, and on it Sol makes about half as many mistakes as its predecessor [20]. That set is selected for having gone wrong, so halving the error count on it is not the same as halving errors across all traffic. OpenAI also reports Sol's coding deception rate at 1.3 percent, down from 10.4 percent, and broken tool disclosure failures falling from 77.5 percent to 4.9 percent [22]. Those are its own evaluations.
Forkast frames the week as the end of a price truce: xAI shipped Grok 4.7 on September 21 at $2 and $6, nine days after Elon Musk endorsed Dario Amodei's call for a coordinated slowdown [14], Anthropic answered with Opus 5.5 at $4 and $20, and OpenAI published ninety minutes later [15]. An antitrust suit filed September 18 alleged the slowdown was an output-restricting cartel, and Forkast argues the price war that followed "makes that argument harder to dismiss" [30]. The order of the releases does not tell you what any of these models cost to serve. OpenAI's stated reason is narrower: improvements in caching and inference "let us serve these models at lower cost" [11].
OpenAI's post also puts a figure on sustained agent use inside the company. Valued at API prices, daily token usage exceeded $600 for the median researcher and $7,000 for researchers at the 90th percentile, and the company says its internal usage "has grown exponentially" [23].
What to watch
- Independent reproduction of the OSWorld 2.0 and DeepSWE v1.1 figures by anyone other than OpenAI.
- Whether OpenAI publishes GPT-5.6 standard rates, so the 50 percent can be checked against a non-promotional baseline.
- Whether the September price moves are cited in the antitrust case filed on September 18.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence55
- Adoption32
- Hype gap+30
- Incentives82
- Confidence62
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
OpenAI introduced GPT-6 Sol and GPT-6 Luna, expanding the GPT-6 family after GPT-6 Astra, which it introduced earlier the same month.
- [2]
OpenAI said it is reducing API prices for Sol and Luna by 50% compared with their GPT-5.6 promotional pricing; the announcement does not state GPT-5.6's standard prices.
- [3]
Forkast reports Sol runs at $2 per million input tokens and $10 per million output tokens.
- [4]
Forkast reports Luna runs at $0.10 per million input tokens and $0.50 per million output tokens.
- [5]
Forkast reports the existing Astra tier at $10 per million input tokens and $50 per million output tokens.
- [6]
Forkast reports that on AutomationBench Sol scores 33.2 percent at $0.27 per task against Opus 5's 26.9 percent at roughly $3 per task, an 11x cost advantage.
- [7]
OpenAI reports that on AutomationBench, GPT-6 Sol at xhigh effort outperforms Claude Opus 5 at max effort at just 9% of Opus 5's cost per task.
- [8]
OpenAI reports that on DeepSWE v1.1, Sol at max effort scores 68.8%, within 1.1 percentage points of Claude Fable 5's highest score in the evaluation (69.9% at xhigh effort), at approximately 80% lower cost per task.
- [9]
OpenAI reports that GPT-6 Luna at max effort scores 66.6% on DeepSWE v1.1 and costs 93% less per task than Opus 5 and 96% less than Fable 5 in those comparisons.
- [10]
Forkast reports that Sol regressed on OSWorld 2.0, the computer-use benchmark, scoring 60.5 percent at xhigh effort versus GPT-5.6 Sol's 65.7 percent.
- [11]
OpenAI wrote that improvements in caching and inference "let us serve these models at lower cost".
- [12]
Forkast reports GPT-6 Sol and Luna arrived on September 22, 2026, and that the pricing was confirmed as permanent, not promotional, by an OpenAI spokesperson speaking to VentureBeat.
- [13]
Forkast reports that Sol matches Grok 4.7's $2 input price.
- [14]
Forkast reports that on September 21 xAI shipped Grok 4.7 at $2/$6, undercutting the $10/$50 frontier by 80 percent, nine days after Elon Musk publicly endorsed Dario Amodei's call for a coordinated slowdown.
- [15]
Forkast reports that on September 22 Anthropic released Opus 5.5 at $4/$20 and OpenAI followed ninety minutes later with Sol at $2/$10 and Luna at $0.10/$0.50.
- [16]
OpenAI reports that on AutomationBench at high effort, GPT-6 Luna improves on its predecessor by 5.4 percentage points at 58% lower cost per task.
- [17]
OpenAI reports that on Agents' Last Exam, GPT-6 Sol at max effort scores 56.4%, above Claude Opus 5's highest score in the evaluation, at 60% lower cost per task.
- [18]
OpenAI says GPT-6 Astra remains the world's best model for computer use, and that "GPT-6 Sol and Luna offer more cost-efficient performance than their predecessors".
- [19]
The reported OSWorld 2.0 score for Sol at xhigh effort is 5.2 percentage points below GPT-5.6 Sol's.
- [20]
OpenAI says its internal factuality evaluation is based on de-identified real-world conversations where users flagged mistakes by its models, and that GPT-6 Sol makes about half as many mistakes as its predecessor on it.
- [21]
OpenAI says GPT-6 Luna at higher effort levels matches GPT-5.6 Sol on factuality at about a hundredth its cost.
- [22]
Forkast reports that OpenAI says GPT-6 Sol's coding deception rate dropped to 1.3 percent from 10.4 percent in GPT-5.6 Sol, and broken tool disclosure failure fell from 77.5 percent to 4.9 percent.
- [23]
OpenAI says that valued at API prices, daily token usage has exceeded $600 for the median researcher and $7,000 for researchers at the 90th percentile, and that "our internal usage has grown exponentially".
- [24]
Sol costs about 67 percent more per output token than Grok 4.7 ($10 against $6).
- [25]
Sol is priced at one fifth of Astra on both input and output tokens, and Luna at one hundredth.
- [26]
If the per-token price fell 50 percent and cost per task fell 58 percent on the same test with an unchanged input-output mix, Luna consumed about 16 percent fewer tokens per task than its predecessor.
- [27]
At Sol's price of $10 per million output tokens, $0.27 per task buys at most about 27,000 output tokens.
- [28]
Forkast wrote that Luna at $0.10/$0.50 "sits below every frontier-class model on the market, including Xiaomi's MiMo-V2.6 Flash at $0.14/$0.28".
- [29]
Luna's output price is about 79 percent above Xiaomi's MiMo-V2.6 Flash ($0.50 against $0.28), while its input price is lower ($0.10 against $0.14).
- [30]
Forkast reports an antitrust lawsuit filed September 18 alleging that the slowdown was an output-restricting cartel, and writes that "the pricing war that followed makes that argument harder to dismiss".
Sources
2 independent publishers whose own reporting we read for this story.
- forkast.newsOpenAI’s GPT-6 Sol and Luna Cut Prices 50% — and the Three-Tier Family Just Ended the Coordinated Slowdown – Forkast
1 article · September 22, 2026
- Introducing GPT-6 Sol and Luna | OpenAI
openai.com
1 article · September 22, 2026