GPT-6.2 Astra Costs 20% Less Than GPT-6 Astra
OpenAI cut standard token rates by twenty percent while keeping the 1,050,000-token context window intact.
GPT-6.2 Astra is OpenAI's most capable model, released on November 17, 2026. The model costs $8.00 per million input tokens and $40.00 per million output tokens, which sits 20 percent below the rates of GPT-6 Astra. It retains the same 1,050,000-token context window as its predecessor while advancing the knowledge cutoff date to May 18, 2026.
OpenAI describes GPT-6.2 Astra as a frontier model built for your hardest projects, from start to finish. The release adds support for subagent delegation through the Multi-agent beta in the Responses Application Programming Interface (API) while dropping cached input rates by 60 percent. OpenAI also lists dated Ultrafast pricing: 3 times the Standard rate through February 28, 2027, and 6 times from March 1, 2027.
What GPT-6.2 Astra Changes From GPT-6 Astra
GPT-6.2 Astra reduces base input and output token rates by 20 percent while retaining the 1,050,000-token context window of GPT-6 Astra. The knowledge cutoff moves forward from April 30, 2026, to May 18, 2026. OpenAI also published a phased pricing schedule for Ultrafast execution, a mode that had no listed pricing on the original GPT-6 Astra rate card.
The operational delta between GPT-6 Astra and GPT-6.2 Astra centers on pricing, retention costs, and inference speed tiers:
| Specification | GPT-6 Astra | GPT-6.2 Astra | Operational Change |
|---|---|---|---|
| Standard Input (per 1M tokens) | $10.00 | $8.00 | 20% price cut |
| Cached Input (per 1M tokens) | $1.00 | $0.40 | 60% price cut |
| Cache Writes (per 1M tokens, 30m) | $12.50 | $10.00 | 20% price cut |
| Standard Output (per 1M tokens) | $50.00 | $40.00 | 20% price cut |
| Long-Context Threshold | Over 272K input tokens | Over 272K input tokens | Unchanged (2x input/cache, 1.5x output) |
| Context Window Limit | 1,050,000 tokens | 1,050,000 tokens | Unchanged total capacity |
| Maximum Input Ceiling | 922,000 tokens | 922,000 tokens | Unchanged maximum prompt space |
| Maximum Output Ceiling | 128,000 tokens | 128,000 tokens | Unchanged maximum generation limit |
| Knowledge Cutoff Date | April 30, 2026 | May 18, 2026 | 18 days later knowledge cutoff |
| Ultrafast Mode Pricing | No pricing published | 3x Standard through Feb 28, 2027; 6x after | Formal multiplier schedule added |
Beyond base token numbers, cached prefix reads dropped from $1.00 down to $0.40 per million tokens. OpenAI also made 12-hour extended cache retention available at $20.00 per million tokens for short-context calls under Standard pricing.
Price, Context and API Limits
Standard pricing for GPT-6.2 Astra is set at $8.00 per million input tokens, $0.40 per million cached input tokens, and $40.00 per million output tokens for standard requests up to 272,000 input tokens. Requests submitted through the Batch endpoint receive a 50 percent discount, while Fast mode runs at twice the Standard rate. Background completion windows are also available in public beta through the Responses API at discounted rates.
For requests containing more than 272,000 input tokens, OpenAI enforces a long-context pricing rule that doubles input and cache rates while raising output rates to 1.5 times the short-context price for the entire request. This surcharge means long-context input tokens cost $16.00 per million, cached input costs $0.80 per million, and output tokens rise to $60.00 per million.
In our internal cost analysis, we modeled how these rates affect production workloads using 1,000 requests each month. For a standard month with 100,000 fresh input tokens and 8,000 output tokens per request without caching, GPT-6.2 Astra costs $1,120.00, compared to $1,400.00 for GPT-6 Astra, saving $280.00 or 20 percent. Our GPT-6 Astra explained guide covers the earlier model. In a cache-heavy month with 20,000 fresh and 80,000 cached input tokens per request, GPT-6.2 Astra totals $512.00 versus $680.00 for GPT-6 Astra, saving $168.00 or 24.7 percent due to the 60 percent cheaper cached rate.
The long-context rule raises the bill once prompts cross the 272,000-token threshold. In our modeled long-context month where every request contains 300,000 fresh input tokens and 8,000 output tokens, GPT-6.2 Astra costs $5,280.00, compared to $6,600.00 on GPT-6 Astra. At headline rates without the long-context rule, those tokens would cost $2,720.00, making the actual bill 94.1 percent higher than headline rates imply. A single request with 272,000 input tokens and 8,000 output tokens costs $2.496, but adding one single token to reach 272,001 input tokens jumps the cost to $4.832.
Standard rate limits scale across developer organization tiers, according to OpenAI's GPT-6.2 Astra model page. Build tier accounts receive 5,000 requests per minute and 1,000,000 tokens per minute. Launch tier accounts receive 10,000 requests per minute and 4,000,000 tokens per minute. Grow tier accounts receive 15,000 requests per minute and 40,000,000 tokens per minute.
- Context window: 1,050,000 total tokens with a 922,000 maximum input ceiling and a 128,000 maximum output ceiling.
- Standard rate card: $8.00 input, $0.40 cached input, $10.00 cache writes (30-minute retention), and $40.00 output per million tokens.
- Long-context threshold: Requests exceeding 272,000 input tokens double input and cache rates and multiply output rates by 1.5 for the full call.
- API interface support: Chat Completions, Responses, and Batch endpoints are supported; Realtime, Assistants, and fine-tuning are unsupported.
Where You Can Use It
GPT-6.2 Astra is accessible through ChatGPT Work, Codex, and the programmatic OpenAI API starting November 17, 2026. Subscriptions on ChatGPT Plus, Pro, Business, Enterprise, and Edu plans include access inside ChatGPT Work and Codex environments. According to an OpenAI Developer Community reply on the thread, the rollout does not add either model to regular Chat.
Enterprise and Edu administrators must first enable GPT-6.2 model access, according to OpenAI's current Learn documentation.
Pro 500 subscribers can use Ultrafast with GPT-6.2 Astra directly inside supported workspaces. Secondary reporting from 9to5Mac states that Ultrafast runs at up to 4.5 times the token rate, consumes more usage allowance, and requires enterprise or education administrators to enable access before end users can select it.
In the API, Ultrafast carries a scheduled pricing multiplier rather than a flat rate. Through February 28, 2027, Ultrafast costs 3 times the Standard rate, pricing GPT-6.2 Astra input at $24.00 and output at $120.00 per million tokens. Starting March 1, 2027, Pacific time, the multiplier increases to 6 times the Standard rate, lifting input to $48.00 and output to $240.00 per million tokens. In our standard month model of 1,000 requests, running Ultrafast costs $3,360.00 through February 2027 and $6,720.00 starting March 2027.
Availability depends on plan, workspace permissions and rollout progress. For the previous generation's quotas, see our guide to GPT-6 Astra usage limits. Developers should update their local Codex command-line interface tools and ChatGPT desktop applications.
Follow the steps to set up GPT-6.2 in Work, Codex or the API.
OpenAI-Reported Results
OpenAI reports that GPT-6.2 Astra establishes new high scores across professional technical benchmarks and scientific research at less than a third of the cost per task of GPT-6 Astra. On the Agents' Last Exam evaluation, OpenAI reports that GPT-6.2 Astra running at extra-high reasoning effort scored 5.2 points above GPT-6 Astra operating at maximum reasoning effort, while cutting cost per task by more than half. These benchmark figures and cost-per-task claims come directly from OpenAI's launch announcement.
OpenAI also reports that GPT-6.2 Astra is its most capable model ever evaluated for cybersecurity tasks, achieving a 22-point gain over GPT-6 Astra on the ExploitBench Internal Port benchmark. Because of the added capability, OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol with tighter safeguards for higher-risk cybersecurity assistance.
OpenAI stated it is working to bring the models to Daybreak. OpenAI says this will allow approved defenders to use the models with reduced refusals.
- Agents' Last Exam: Scored 5.2 points higher at extra-high reasoning effort than GPT-6 Astra at maximum effort, at less than half the task cost (OpenAI-reported).
- Professional and scientific evals: Sets new highs across professional work and scientific research at less than a third of the cost per task of GPT-6 Astra (OpenAI-reported).
- ExploitBench Internal Port: Jumps 22 points over GPT-6 Astra (OpenAI-reported).
- Reasoning controls: The API requires reasoning tokens and automatically rejects any call that attempts to set reasoning effort to none.
Who Should Choose Astra Over Sol
Choose GPT-6.2 Astra when your workload involves the hardest coding and research projects where its cost per task claim matters. The model builds and revises software, works with documents, and supports computer use in the Responses API. Our comparison of GPT-6.1 Sol vs GPT-6 Astra shows how previous flagship and efficient tiers divided similar engineering workloads.
GPT-6.2 Astra is not the right choice for teams whose everyday work GPT-6.2 Sol covers. GPT-6.2 Sol is priced at $2.00 per million input tokens and $10.00 per million output tokens, which is exactly 25 percent of GPT-6.2 Astra's operational price. In our modeled standard month of 1,000 requests, GPT-6.2 Sol costs $280.00 compared to $1,120.00 for GPT-6.2 Astra. Furthermore, teams needing conversational access in the standard ChatGPT Chat interface cannot use Astra, as the rollout does not add it to regular Chat.
Our recommendation would change if independent results disagree with OpenAI's reported cost-per-task efficiency. To budget your migration accurately, calculate your average prompt token depth to verify whether requests trigger the 272,000-token long-context rule before routing high-volume traffic to the gpt-6.2-astra endpoint.
You can also compare GPT-6.2 Astra with Claude Fable 5.2 on current rates, access and documented workflows.
Frequently Asked Questions
- GPT-6.2 Astra is OpenAI's most capable reasoning model, released on November 17, 2026, for complex coding, computer interaction, and professional research tasks.
- GPT-6.2 Astra costs 20 percent less per standard input and output token, cuts cached input prices by 60 percent, advances the training cutoff date to May 18, 2026, and adds a scheduled pricing multiplier for Ultrafast mode.
- Standard API rates for GPT-6.2 Astra are $8.00 per million input tokens, $0.40 per million cached input tokens, and $40.00 per million output tokens for requests up to 272,000 input tokens. Requests over 272,000 input tokens double input and cache rates and multiply output rates by 1.5.
- No. A reply on OpenAI's launch thread says the rollout does not add either model to regular Chat. GPT-6.2 Astra is in ChatGPT Work, Codex and the API.
- GPT-6.2 Astra features a 1,050,000-token context window with a 922,000 maximum input limit, a 128,000 maximum output limit, and a knowledge cutoff date of May 18, 2026.
- Pick GPT-6.2 Astra for your hardest coding and research projects where its cost per task claim matters. Choose GPT-6.2 Sol when everyday work covers the task, since its list price per token is 75 percent lower.
Plan Where GPT-6.2 Astra Fits Your Work
At Layer3Labs, we build custom agents and workflow automation for small and mid-sized businesses. Book an AI workflow audit to map where GPT-6.2 Astra fits your work.
Book a Consultation