GeethanTechGeethanTechTech, decoded daily.
LatestAISoftware & PlatformsCybersecurityStartups & VentureTech Business & Markets
All posts
AI3 months ago

OpenAI Lowers GPT-5.6 Sol API Prices Through at Least November 21

GPT-5.6 Sol now costs $4 per million input tokens and $20 per million output tokens under promotional pricing available at least through November 21, 2026. Prompts above 272K input tokens use higher rates for the full request.

OpenAI Lowers GPT-5.6 Sol API Prices Through at Least November 21

OpenAI’s GPT-5.6 Sol model page now lists promotional API prices of $4 per million input tokens, $0.40 per million cached input tokens, and $20 per million output tokens. OpenAI says these rates will remain available at least through November 21, 2026.

The pricing applies to applications that call gpt-5.6-sol and to those using the gpt-5.6 alias, which routes requests to Sol. It is not a performance announcement, and OpenAI has not published the prices that will follow the promotional window.

The base rates are promotional

OpenAI describes the listed prices as a 20% reduction in input pricing and a 33% reduction in output pricing. Those percentages are the company’s comparison, not an independent assessment of model value or application cost.

The cached-input rate is $0.40 per million tokens. Cache writes are billed at 1.25 times the uncached input-token rate, so a useful cost model needs to distinguish uncached input, cached reads, cache writes, and output. The two headline rates alone do not describe every billable path.

OpenAI says the promotion is available at least through November 21. That establishes a minimum availability window, not a confirmed end date or a future price.

Large-context requests cross a separate threshold

A different schedule applies when a prompt contains more than 272K input tokens. OpenAI says input is then priced at twice the listed rate and output at 1.5 times the listed rate for the full request, not only for the tokens above the threshold.

Based on the published multipliers, that makes the uncached input rate $8 per million tokens and the output rate $30 per million tokens for those requests. The model page does not provide a worked billing example, so teams using both long contexts and prompt caching should validate estimates against their actual usage records.

The threshold matters because GPT-5.6 Sol supports a 1,050,000-token context window, with up to 922,000 input tokens and 128,000 output tokens. A request can therefore remain within the model’s limits while falling into the higher pricing schedule.

The model alias belongs in cost reviews

OpenAI states that the gpt-5.6 alias routes to gpt-5.6-sol. Teams already using the alias are consuming Sol under that model name even when their configuration does not contain the suffixed identifier.

Cost reviews should therefore inventory model aliases as well as explicit model IDs across application code, gateways, evaluation systems, and background jobs. They should also separate ordinary requests from prompts above 272K input tokens before comparing spend.

The pricing update does not establish a change in model quality, latency, reliability, or benchmark performance. Those questions require separate evidence.

Budgeting needs a post-promotion scenario

For the current window, teams can model standard and large-context requests separately, then break each group into uncached input, cached input, cache writes, and output.

Forecasts extending beyond November 21 need a separate assumption. OpenAI may continue the promotion, replace it, or adopt another schedule, but the current documentation does not say which outcome will occur.

The practical change is a lower published base price with two important boundaries: the long-context multiplier begins above 272K input tokens, and the price after the stated promotional window remains unknown.

Continue with this topicAI
Browse categoryarrow_forward
Continue reading

Related posts

A brass cylinder above three circular impressions in clay.
AIabout 24 hours agoBy megan

Inkling 2 Mini brings a looped AI model to local hardware

Thinking Machines Lab has released open weights for a 26.4-billion-parameter model that runs its middle layers three times per token. Local inference is supported on Apple silicon systems with at least 32GB of memory, but the company warns of limits in factual recall and agent reliability.

Read article
Conceptual illustration of a human Mac workspace beside a separate abstract task window with document, app, and web shapes.
AI2 days agoBy megan

Gemini for Mac adds computer use in a limited US rollout

Google says Gemini can work across Mac apps, local files, and websites in a separate background window. The English-language rollout is limited to eligible US personal accounts on paid plans.

Read article
GeethanTech
GeethanTech
Tech, decoded daily.
Read
Latest postsAISoftware & PlatformsCybersecurityStartups & VentureTech Business & MarketsRSS feed
Connect
mailEmaillanguageWebsiteinLinkedInIGInstagramfFacebook@Xsmart_displayYouTube
From the publisher

Read all our publications: GeethanPost | GeethanTech

A chapter in the ElegantHumanity story

Powered byElegantArc