By Jonathan West · Updated Nov 19, 2026

GPT-6.2 Pricing Puts Astra at $8/$40 and Sol at $2/$10

A breakdown of standard token fees, the 272K long-context doubling rule, Ultrafast schedules, and ChatGPT workspace tiers.

By Jonathan West · Updated Nov 19, 2026

GPT-6.2 Astra costs $8.00 per million input tokens and $40.00 per million output tokens, while GPT-6.2 Sol costs $2.00 per million input tokens and $10.00 per million output tokens. Prompt caching lowers input costs to $0.40 per million tokens for GPT-6.2 Astra and $0.10 per million tokens for GPT-6.2 Sol. When an API request exceeds 272,000 input tokens, input and cache rates double, and output rates increase by 50 percent for the entire request.

Ultrafast mode runs at three times the standard token rate through February 28, 2027, before increasing to six times standard rates on March 1, 2027. Both models are available through the Application Programming Interface (API) and in ChatGPT Work and Codex. A launch-thread reply says the rollout does not add them to regular ChatGPT Chat.

Headline standard token rates for GPT-6.2 do not tell the whole pricing story. While OpenAI launched both GPT-6.2 Astra and GPT-6.2 Sol without Flex processing, eligible background requests can reduce model token pricing through completion windows.


GPT-6.2 API Prices Per Million Tokens

Standard API rates for GPT-6.2 Astra are $8.00 per million input tokens and $40.00 per million output tokens, while GPT-6.2 Sol rates are $2.00 per million input tokens and $10.00 per million output tokens. By our arithmetic on the rates in OpenAI's launch announcement, GPT-6.2 Sol costs exactly one quarter of GPT-6.2 Astra per token, on both input and output. For rates on other models, see our AI model pricing registry.

Rate TypeGPT-6.2 AstraGPT-6.2 SolGPT-6 AstraGPT-6.1 Sol
Input (per 1M tokens)$8.00$2.00$10.00$2.00
Cached Input (per 1M tokens)$0.40$0.10$1.00$0.10
Cache Writes (30 min)$10.00$2.50$12.50$2.50
Output (per 1M tokens)$40.00$10.00$50.00$10.00

Our calculations show GPT-6.2 Astra lowers standard input, output, and cache-write rates by 20 percent compared to the prior-generation GPT-6 Astra, while cached input rates drop by 60 percent. GPT-6.2 Sol holds the same headline token rates as GPT-6.1 Sol. Both models reject API requests that set reasoning effort to none.

Compare GPT-6.2 Astra versus Claude Fable 5.2 for equal standard base rates and different access and cache conditions.

  • Standard input rates: $8.00 per million tokens for Astra, $2.00 per million tokens for Sol.
  • Cached input discounts: $0.40 per million tokens for Astra, $0.10 per million tokens for Sol.
  • Standard output rates: $40.00 per million tokens for Astra, $10.00 per million tokens for Sol.
  • Batch API discounts: 50 percent off standard rates for both models, with Fast mode set at twice the standard rate.

Discounted Background Completion Windows

Eligible Standard background requests reduce model token rates by 30% for a 2-hour window or 40% for an 8-hour window. The feature is in public beta for the Responses application programming interface (API) across GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra, and GPT-6 Luna. Workloads using the default soonest setting receive no window discount.

Qualifying for these lower rates requires specific request parameters. Calls must set background=true and configure background_options.completion_window to 2h or 8h. The request must use Standard processing, which you select by setting service_tier=default (the auto setting qualifies only when it resolves to Standard). Streaming must remain false or unset for 2h or 8h completion windows.

The completion window sets a processing target for the initial response starting when OpenAI accepts the request. That target does not guarantee a completion deadline for a whole multi-step workflow. Follow-up requests can also qualify for the discount, and requests that share a continuation identifier, window, model, and project receive higher queue priority.

Processing SelectionCompletion Window TargetAstra Input (per 1M)Astra Output (per 1M)Sol Input (per 1M)Sol Output (per 1M)
Standard (soonest)Not published$8.00$40.00$2.00$10.00
Standard (2h)2 hours$5.60$28.00$1.40$7.00
Standard (8h)8 hours$4.80$24.00$1.20$6.00

Technical limitations apply to caching, shared limits, and data retention. Requests in the 2h and 8h windows share rate limits with OpenAI Batch processing, and input tokens count against limits at submission time. Standard requests outside discounted completion windows cannot reuse prompt cache entries written by 2h or 8h requests, even when prompts match character for character. Background processing also temporarily stores response data encrypted at rest, even for organizations using Zero Data Retention (ZDR) or passing store=false.

At Layer3Labs, in our internal cost analysis, we modeled monthly token expenses for a team running 1,000 short-context requests to evaluate the practical impact of completion windows. The reproducible scenario assumes each request processes 100,000 uncached input tokens and generates 8,000 output tokens. Under Standard rates with the default soonest window, the baseline monthly cost equals $1,120 for Astra and $280 for Sol. Applying the 2h window lowers those token totals to $784 for Astra and $196 for Sol. Moving to the 8h window reduces them further to $672 for Astra and $168 for Sol. These calculated figures reflect base token rates and exclude non-token expenses such as tools, storage, cache writes, data residency, or taxes.


Cached and Long-Context Rates

Requests with more than 272,000 input tokens double standard input and caching rates and incur a 50 percent surcharge on output tokens for the entire call. This long-context rule applies to both GPT-6.2 Astra and GPT-6.2 Sol. For context-heavy pipelines, passing this threshold changes the cost of every query.

Our calculation of the single-request boundary on GPT-6.2 Astra illustrates this cost cliff. A request containing 272,000 fresh input tokens and 8,000 output tokens costs $2.496 under standard rates. Going just past 272,000 input tokens (we read 272K as 272,000) doubles the input rate to $16.00 per million and raises the output rate to $60.00 per million, pushing the total cost of that single request to about $4.83.

Prompt caching reduces costs for cached input tokens. Cache writes covering 30 minutes of retention cost $10.00 per million tokens on GPT-6.2 Astra and $2.50 per million tokens on GPT-6.2 Sol. For workloads that need 12-hour retention windows, short-context requests incur cache-write fees of $20.00 per million tokens on Astra and $5.00 per million tokens on Sol.


Ultrafast Pricing

Ultrafast mode costs three times the standard rate through February 28, 2027, and six times the standard rate starting March 1, 2027, Pacific time. This dated schedule applies to both OpenAI's GPT-6.2 Astra and GPT-6.2 Sol models. GPT-6.1 Sol lists a flat six-times multiplier.

Secondary reporting from 9to5Mac reports that Ultrafast runs up to 4.5 times the token rate and uses more of a subscription's allowance.

On GPT-6.2 Astra, the three-times rate sets Ultrafast at $24.00 per million input tokens and $120.00 per million output tokens through February 28, 2027. From March 1, 2027, the six-times multiplier puts Astra Ultrafast at $48.00 per million input tokens and $240.00 per million output tokens. GPT-6.2 Sol Ultrafast costs $6.00 input and $30.00 output through February 28, 2027, then $12.00 input and $60.00 output from March 1, 2027.


Which ChatGPT Plans Include Them

GPT-6.2 Astra and GPT-6.2 Sol are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu subscribers. An OpenAI Developer Community post clarifies that this deployment excludes regular ChatGPT Chat interfaces. Availability depends on plan type, organizational workspace permissions, and rollout progress.

Enterprise and Edu administrators must first enable GPT-6.2 model access, according to OpenAI's current Learn documentation.

Subscription PlanGPT-6.2 Astra AccessGPT-6.2 Sol AccessUltrafast Support
PlusChatGPT Work and CodexChatGPT Work and CodexNot published
BusinessChatGPT Work and CodexChatGPT Work and CodexNo Business entitlement stated in cited Learn update
Pro 200ChatGPT Work and CodexChatGPT Work and CodexIncluded, starting with GPT-6.2 Sol
Pro 500ChatGPT Work and CodexChatGPT Work and CodexIncluded with either model
EnterpriseAfter admin model enablementAfter admin model enablementOnce an admin enables it (per 9to5Mac)
EduAfter admin model enablementAfter admin model enablementOnce an admin enables it (per 9to5Mac)

OpenAI brought Ultrafast to Pro 200 subscribers for the first time, starting with GPT-6.2 Sol. Pro 500 accounts can use Ultrafast with either model, while 9to5Mac reports that Enterprise and Edu accounts gain access once a workspace administrator enables the feature. Teams looking at historical rate differences can check GPT-6 Astra pricing and GPT-6.1 Sol pricing to compare prior-generation rates.

After checking plan and workspace access, activate GPT-6.2 in Work, Codex or the API.


Cost Example

In our internal cost analysis, we modeled five monthly workload scenarios across 1,000 API requests to quantify real-world billing differences between models. Our assumptions evaluate standard published rate cards without Batch discounts, Flex processing, Fast premiums, regional adjustments, or cache-write storage fees. We apply the 272,000-token boundary per request. The Ultrafast scenarios below use short-context rates.

Workload ScenarioGPT-6.2 AstraGPT-6 AstraGPT-6.2 SolGPT-6.1 Sol
Standard Month (100K fresh input, 8K output)$1,120.00$1,400.00$280.00$280.00
Cache-Heavy Month (20K fresh, 80K cached, 8K output)$512.00$680.00$128.00$128.00
Long-Context Month (300K fresh input, 8K output)$5,280.00$6,600.00$1,320.00$1,320.00
Ultrafast Month Through Feb 28, 2027 (100K fresh, 8K output)$3,360.00$8,400.00$840.00$1,680.00
Ultrafast Month From Mar 1, 2027 (100K fresh, 8K output)$6,720.00$8,400.00$1,680.00$1,680.00

Our calculations show GPT-6.2 Astra saves 20.0 percent over GPT-6 Astra in standard months ($1,120.00 versus $1,400.00) and 24.7 percent in cache-heavy months ($512.00 versus $680.00) due to cheaper cached tokens. In the long-context month, 300,000 fresh input tokens push every request over the 272K threshold. That rule pushes GPT-6.2 Astra spend to $5,280.00, which is 94.1 percent higher than headline rates imply. Sol costs 25 percent of Astra across all equivalent workloads.

OpenAI reports that GPT-6.2 Astra completes scientific and professional tasks at less than one-third the cost per task of GPT-6 Astra, and scores 5.2 points higher on Agents' Last Exam at xhigh reasoning effort at less than half the cost per task. On FrontierCode 2.0, OpenAI reports GPT-6.2 Sol nearly matches Astra on merge-ready changes at roughly one-fifth the operational cost.


Our Takeaway

Choose the 2-hour or 8-hour background completion windows on OpenAI only when your application can tolerate delayed initial responses and your code satisfies every Standard processing requirement without streaming. In our internal cost analysis at Layer3Labs, moving a 1,000-request short-context workload to an 8-hour window dropped Astra token costs from $1,120 to $672 and Sol from $280 to $168. Defaulting to soonest eliminates those savings completely, and separate costs for tool calls, storage, and prompt cache boundaries mean total invoices will not drop by an unconditional 30% to 40%. Teams handling independent workloads supplied upfront should evaluate Batch processing separately rather than treating background completion windows as a third service tier.

To control monthly API expenses, check your prompt sizes to keep inputs below the 272,000-token threshold before deploying GPT-6.2 pricing models into production.

Frequently Asked Questions

  • GPT-6.2 Sol costs $2.00 per million input tokens, $0.10 per million cached input tokens, and $10.00 per million output tokens under standard pricing. Cache writes covering 30 minutes cost $2.50 per million tokens.
  • GPT-6.2 Astra costs $8.00 per million input tokens, $0.40 per million cached input tokens, and $40.00 per million output tokens under standard pricing. Cache writes covering 30 minutes cost $10.00 per million tokens.
  • GPT-6.2 Sol has identical headline rates to GPT-6.1 Sol at $2.00 per million input and $10.00 per million output tokens. On Ultrafast, GPT-6.2 Sol costs 3x Standard through February 28, 2027, against a flat 6x on GPT-6.1 Sol. From March 1, 2027 both are 6x.
  • GPT-6.2 Ultrafast costs three times standard rates through February 28, 2027, and six times standard rates starting March 1, 2027, Pacific time. Through February 28, 2027, Astra Ultrafast costs $24.00 input and $120.00 output, while Sol Ultrafast costs $6.00 input and $30.00 output per million tokens.
  • Requests containing more than 272,000 input tokens double input and cache rates while increasing output rates by 50 percent for the entire request. For example, a GPT-6.2 Astra call just over 272,000 input tokens with 8,000 output tokens costs about $4.83, against $2.496 at 272,000 tokens.
  • GPT-6.2 Astra and Sol are available in ChatGPT Work and Codex for Plus, Pro 200, Pro 500, Business, Enterprise, and Edu accounts. Ultrafast mode is supported on Pro 500 for both models and on Pro 200 starting with Sol, while Enterprise and Edu tiers require administrator enablement for model access as well as Ultrafast, subject to rollout.
  • Discounted background completion windows reduce base token rates by 30% for a 2-hour window and 40% for an 8-hour window on eligible Standard Responses API requests. On GPT-6.2 Astra, rates drop from the Standard $8 input and $40 output per million tokens to $5.60/$28 in the 2-hour window and $4.80/$24 in the 8-hour window. On GPT-6.2 Sol, rates fall from $2 input and $10 output to $1.40/$7 for 2 hours and $1.20/$6 for 8 hours. The default soonest setting provides no rate discount, and non-token expenses like tools and storage remain separate.

Plan Your GPT-6.2 Costs

At Layer3Labs, we build custom agents and workflow automation for small and mid-sized businesses. Book an AI workflow audit to plan your GPT-6.2 costs.

Book a Consultation