GPT-6.2 Astra and Sol Guide: API Pricing, Specs, and Which to Pick (2026)

What Is GPT-6.2: Astra and Sol, Released November 17
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol together on November 17, 2026. Unlike the GPT-6 family's three-wave rollout in September, both models were replaced at once: Astra takes over the top slot from GPT-6 Astra, and Sol takes over the workhorse slot from GPT-6.1 Sol.
How the official model pages describe them:
- GPT-6.2 Astra: "Our most capable model for your hardest projects, from start to finish." It can build and revise software, use a computer, and work with documents
- GPT-6.2 Sol: "Our go-to model for everyday work, from coding to powering agents at scale."
Current GPT-6 Family Lineup and Pricing
| Model | Model ID | Input | Cached input | Cache write (30 min) | Output |
|---|---|---|---|---|---|
| GPT-6.2 Astra | gpt-6.2-astra | $8 | $0.40 | $10 | $40 |
| GPT-6.2 Sol | gpt-6.2-sol | $2 | $0.10 | $2.50 | $10 |
| GPT-6 Astra | gpt-6-astra | $10 | $1 | $12.50 | $50 |
| GPT-6.1 Sol | gpt-6.1-sol | $2 | $0.10 | $2.50 | $10 |
| GPT-6.1 Terra | gpt-6.1-terra | $1 | $0.05 | $1.25 | $5 |
| GPT-6 Luna | gpt-6-luna | $0.10 | $0.01 | $0.125 | $0.50 |
Per million tokens, USD, Standard tier, for requests with 272K input tokens or fewer. Sources: OpenAI developer docs model pages for GPT-6.2 Astra and GPT-6.2 Sol, and the official pricing page (checked 2026-11-19). GPT-6 Astra and GPT-6.1 Sol are still listed on the official pricing page.
Pricing: Astra Down 20%, Sol Unchanged
Line-by-Line Comparison with the Previous Generation
| Item (per million tokens) | GPT-6.2 Astra | GPT-6 Astra | GPT-6.2 Sol | GPT-6.1 Sol |
|---|---|---|---|---|
| Input | $8 | $10 | $2 | $2 |
| Cached input | $0.40 | $1 | $0.10 | $0.10 |
| Cache write (30 min) | $10 | $12.50 | $2.50 | $2.50 |
| Cache write (12 hours) | $20 | $25 | $5 | $5 |
| Output | $40 | $50 | $10 | $10 |
| Over 272K input: input/output | $16/$60 | $20/$75 | $4/$15 | $4/$15 |
| Batch: input/output | $4/$20 | $5/$25 | $1/$5 | $1/$5 |
Worked out from the two sets of official prices: GPT-6.2 Astra's input, output, and cache write prices are 80% of GPT-6 Astra's, and its cached input is 40%. Every cell for GPT-6.2 Sol matches GPT-6.1 Sol, so switching does not change your bill.
Billing Rules Shared by Both Models
- Long-prompt surcharge: OpenAI splits at 272K input tokens per request; above that, input and cache rates double and output costs 1.5 times as much, and it applies to the whole request, not just the tokens above the threshold
- Cache writes are billed: a write with 30-minute retention costs 1.25 times the uncached input price (Astra $10, Sol $2.50); 12-hour retention for short-context Standard requests is Astra $20, Sol $5
- Batch is half price: in OpenAI's words, "Batch costs 50% of Standard rates"
- Fast mode is 2x: twice the Standard rate for the same context length (OpenAI renamed Priority processing to Fast mode on 2026-07-30)
- Ultrafast has two phases: 3 times the Standard rate through 2027-02-28, then 6 times from 2027-03-01 (Pacific time)
- Background request discount (public beta): eligible background requests in the Responses API cost 30% below Standard for the 2-hour completion window and 40% below for the 8-hour window
- Regional processing adds 10%: data residency endpoints carry a 10% uplift for models released on or after 2026-03-05
One more thing to watch: as of 2026-11-19, the Flex table on the official pricing page does not list GPT-6.2 Astra or GPT-6.2 Sol (GPT-6 Astra and GPT-6.1 Sol are both there). If you run GPT-6.1 Sol on Flex today, confirm how you'll be billed before moving to GPT-6.2 Sol.
Key Specs: How Astra and Sol Differ
| Item | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|
| Official positioning | Most capable, for the hardest projects | Go-to for everyday work, coding to agents at scale |
| Context | 1,050,000 tokens | 1,050,000 tokens |
| Max input | 922,000 | 922,000 |
| Max output | 128,000 | 128,000 |
| Knowledge cutoff | 2026-05-18 | 2026-06-01 |
| Input/output | Text, image/text | Text, image/text |
reasoning.effort set to none | Rejected | Rejected |
| Price (input/output) | $8/$40 | $2/$10 |
The other official specs are the same for both:
- Endpoints: Chat Completions, Responses, and Batch are supported; Realtime, Assistants, fine-tuning, and image and speech generation endpoints are not
- Features: streaming, structured outputs, function calling, file search, image input, web search, and prompt caching are supported; fine-tuning and predicted outputs are not
- Responses API tools: web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP, and tool search; the Multi-agent beta can also delegate work to subagents
- Rate limits (Standard): Build 5,000 RPM / 1M TPM, Launch 10,000 RPM / 4M TPM, Grow 15,000 RPM / 40M TPM
The knowledge cutoff is worth noting: Sol's is about two weeks later than Astra's (2026-06-01 versus 2026-05-18), so for questions involving late May to early June, the cheaper Sol actually knows more.
Cost Math (Arithmetic, Not Benchmarks)
Everything below is plain arithmetic on official unit prices; it says nothing about model quality.
Scenario A: A Typical Workload
Assumptions: 50 million input tokens and 10 million output tokens per month, every request under 272K, no caching, Standard tier.
| Model | Input cost | Output cost | Monthly total |
|---|---|---|---|
| GPT-6.2 Astra | 50M × $8 = $400 | 10M × $40 = $400 | $800 |
| GPT-6 Astra | 50M × $10 = $500 | 10M × $50 = $500 | $1,000 |
| GPT-6.2 Sol | 50M × $2 = $100 | 10M × $10 = $100 | $200 |
| GPT-6.1 Sol | 50M × $2 = $100 | 10M × $10 = $100 | $200 |
| (Reference) GPT-5.6 Sol promotional price | 50M × $4 = $200 | 10M × $20 = $200 | $400 |
For the same workload, GPT-6.2 Astra costs 4 times as much as GPT-6.2 Sol. GPT-5.6 Sol's $4/$20 is a promotional price that OpenAI says runs until 2026-11-30; the price after that has not been published.
The Same Workload in Other Billing Modes
| Billing mode | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|
| Standard | $800 | $200 |
| Batch (half price) | $400 | $100 |
| Background request, 8-hour window (40% off) | $480 | $120 |
| Background request, 2-hour window (30% off) | $560 | $140 |
| Fast mode (2x) | $1,600 | $400 |
| Ultrafast, through 2027-02-28 (3x) | $2,400 | $600 |
| Ultrafast, from 2027-03-01 (6x) | $4,800 | $1,200 |
Ultrafast pricing will double on 2027-03-01. If you turn on Ultrafast for speed today, budget for the post-March rate now.
Scenario B: Long Prompts (Over 272K per Request)
Assumptions: 1,000 requests per month, each with 300,000 input and 5,000 output tokens, for 300 million input and 5 million output tokens per month, no caching. At 300K each, every request is over the 272K threshold, so the whole request is billed at long-prompt rates.
| Method | Input cost | Output cost | Monthly total |
|---|---|---|---|
| GPT-6.2 Astra (wrong: using short-prompt rates) | 300M × $8 = $2,400 | 5M × $40 = $200 | $2,600 |
| GPT-6.2 Astra (actual: long-prompt rates) | 300M × $16 = $4,800 | 5M × $60 = $300 | $5,100 |
| GPT-6.2 Sol (wrong: using short-prompt rates) | 300M × $2 = $600 | 5M × $10 = $50 | $650 |
| GPT-6.2 Sol (actual: long-prompt rates) | 300M × $4 = $1,200 | 5M × $15 = $75 | $1,275 |
| (Reference) GPT-6 Astra, long-prompt rates | 300M × $20 = $6,000 | 5M × $75 = $375 | $6,375 |
Budgeting with short-prompt rates underestimates the bill by almost half. If you can keep each request under 272K input tokens (for example, by splitting documents or sending only the relevant sections), the same token volume goes back to short-prompt rates; whether splitting adds extra tokens is something to measure on your own data.
Scenario C: A Long, Repeated System Prompt with Caching
Assumptions: same as Scenario A, with 40 million of the 50 million input tokens served from cache. This is the best case and excludes cache write costs.
| Model | Cached input | Uncached input | Output | Monthly total |
|---|---|---|---|---|
| GPT-6.2 Astra | 40M × $0.40 = $16 | 10M × $8 = $80 | $400 | $496 |
| GPT-6 Astra | 40M × $1 = $40 | 10M × $10 = $100 | $500 | $640 |
| GPT-6.2 Sol | 40M × $0.10 = $4 | 10M × $2 = $20 | $100 | $124 |
Cache writes are billed separately (Astra's 30-minute retention costs $10 per million, 1.25 times uncached input). If your requests are often more than 30 minutes apart and the cache keeps expiring and being rewritten, the real cost will be higher than the table above. Estimate your hit rate before turning caching on, or consider 12-hour retention.
Switching from GPT-6 Astra, GPT-6.1 Sol, or GPT-5.6: What to Change
1. Model ID
Change gpt-6-astra to gpt-6.2-astra and gpt-6.1-sol to gpt-6.2-sol. The official model pages do not currently list any dated snapshots. Keep the model name in a config file or environment variable so the next upgrade is a one-line change.
2. reasoning.effort Can't Be none
Both official model pages say: "The API rejects requests that set reasoning effort to none." This affects code coming from GPT-5.6, GPT-6.1 Terra, GPT-6 Luna, or GPT-6 Sol: if it sets effort to none to save money or time, simply swapping the model name will get the request rejected; switch to another effort value. GPT-6 Astra and GPT-6.1 Sol never accepted none, so moving from those two is unaffected.
3. If You Use Flex, Check Billing First
As of 2026-11-19, the Flex table on the official pricing page does not include GPT-6.2 Astra or Sol. Batch jobs that run on Flex with GPT-6 Astra or GPT-6.1 Sol can consider Batch (the model pages state Batch is half price) or the Responses API's background completion windows.
4. The 272K Threshold and the Ultrafast Date
The long-prompt surcharge works the same way as in the GPT-6 family, but the Ultrafast multiplier will rise from 3x to 6x on 2027-03-01. Build both into your budget and alert thresholds.
5. Coming from GPT-5.6 Sol
GPT-5.6 Sol's promotional price of $4/$0.40/$20 runs until 2026-11-30, and OpenAI has not published the price after that. GPT-6.2 Sol is $2/$10, half even compared with the promotional price. Also, gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3, and o3-pro will shut down on 2026-12-11 (OpenAI recommends gpt-6.1-sol, with gpt-6.1-terra for mini and gpt-6-luna for nano; see OpenAI deprecations); teams still on those models can widen their evaluation to include GPT-6.2.
6. Reasoning Preservation and Top-Level instructions Changed
OpenAI's API changelog (2026-11-17) states: "GPT-6.2 changes how the Responses API preserves reasoning when you edit a conversation's prefix and how it handles the top-level instructions field." If your integration rewrites earlier messages or relies on the top-level instructions field, check OpenAI's GPT-6.2 migration guide and run another round of tests before switching models.
CloudInsight Enterprise OpenAI API Procurement | Local Invoices, No Foreign Credit Card Needed
GPT-6.2's unit prices have come down, but the 272K threshold and the Fast and Ultrafast multipliers make monthly bills harder to predict. What enterprises get stuck on most is often not the technology but payment, invoicing, and budget control.
Contact CloudInsight for an OpenAI API Enterprise Quote
We offer one-stop procurement for Claude, GPT, and Gemini, Taiwan uniform invoices, and technical support in Chinese.
How to Choose: GPT-6.2 Astra, GPT-6.2 Sol, and the Previous Generation
| Workload | Recommendation | Why |
|---|---|---|
| Everyday coding, agent workflows, professional documents | GPT-6.2 Sol | The official everyday pick; same price as GPT-6.1 Sol with a newer knowledge cutoff |
| The hardest reasoning, large end-to-end projects | GPT-6.2 Astra | The official most capable model, 20% cheaper per token than GPT-6 Astra |
| Already on GPT-6 Astra | Evaluate moving to GPT-6.2 Astra | About 80% of the bill for the same workload; more savings when caching is heavy |
| Already on GPT-6.1 Sol via Flex | Check billing before switching | The pricing page's Flex table doesn't list GPT-6.2 yet |
| Long documents that often exceed 272K input per request | Do the math first | The whole request's input doubles and output is 1.5x; Astra in Scenario B goes from $2,600 to $5,100 |
| High-volume, simple summarization, classification, routing | GPT-6 Luna or GPT-6.1 Terra | Terra is half of GPT-6.2 Sol's unit price and Luna is one twentieth; move up only if quality falls short |
This guide does not cite benchmarks. Our advice: run 50 to 100 real cases through both GPT-6.2 Sol and GPT-6.2 Astra, then decide whether the 4x price difference is worth it.
FAQ: GPT-6.2
How much do GPT-6.2 Astra and Sol cost?
Per million tokens: Astra $8 input, $0.40 cached input, $10 cache write, $40 output; Sol $2 input, $0.10 cached input, $2.50 cache write, $10 output. Batch is half price.
When was GPT-6.2 released?
GPT-6.2 Astra and GPT-6.2 Sol were both released on 2026-11-17.
How much cheaper is GPT-6.2 Astra than GPT-6 Astra?
Input, output, and cache writes are 80% of GPT-6 Astra's prices ($8/$40 versus $10/$50), and cached input is 40% ($0.40 versus $1). For the same 50 million input and 10 million output tokens a month, the bill drops from $1,000 to $800.
Do GPT-6.2 Sol and GPT-6.1 Sol cost the same?
Yes. Their Standard, long-prompt, and Batch prices are identical on the official pricing page. The differences: GPT-6.2 Sol is not in the Flex table yet, and its knowledge cutoff is newer (2026-06-01 vs. 2026-04-30).
Does GPT-6.2's 1M context cost extra?
The context window is 1,050,000 tokens, with up to 922,000 input tokens per request. Above 272K input tokens, the whole request's input and cache rates double and output costs 1.5 times as much.
How much does Ultrafast cost?
For GPT-6.2 Astra and Sol, Ultrafast is 3 times the Standard rate through 2027-02-28 and 6 times from 2027-03-01 (Pacific time). Fast mode is 2x.
Conclusion: Default to Sol, Save Astra for the Hardest Work
Three Core Recommendations
- Move everyday work to GPT-6.2 Sol now: same price as GPT-6.1 Sol with a newer knowledge cutoff; if you are coming from GPT-5.6, GPT-6.1 Terra, GPT-6 Luna, or GPT-6 Sol, first remove any
noneeffort setting, and check billing for Flex workloads - If you're on GPT-6 Astra, evaluate GPT-6.2 Astra: unit prices are 20% lower, and cached input is down to 40%
- Do the math on long prompts and Ultrafast: over 272K, the whole request is surcharged, and Ultrafast will rise from 3x to 6x on 2027-03-01
Ready to Adopt GPT-6.2?
Contact the CloudInsight sales team and we'll help plan your GPT-6.2 Astra and Sol usage mix and handle payment and invoicing.
Join our LINE Official Account for instant model-selection consultation.
Further Reading
- GPT-6 Astra, 6.1 Sol, Luna: API Pricing, Specs, and Which to Pick
- GPT-5.6 Sol, Terra, Luna: Which Tier Should You Pick?
- OpenAI Model Shutdown Schedule (Late 2026 to 2027)
- OpenAI API Pricing Explained
- LLM API Cost Optimization
References
- GPT-6.2 Astra model page — OpenAI developer docs (checked 2026-11-19)
- GPT-6.2 Sol model page — OpenAI developer docs (checked 2026-11-19)
- OpenAI API pricing (checked 2026-11-19)
- Introducing GPT-6.2 Astra and Sol — OpenAI (2026-11-17)
- Deprecations — OpenAI developer docs
- API changelog — OpenAI developer docs (checked 2026-11-19)
Need Professional Cloud Advice?
Whether you're evaluating cloud platforms, optimizing existing architecture, or looking for cost-saving solutions, we can help
Book Free ConsultationRelated Articles
GPT-6 Astra, 6.1 Sol, Luna: API Pricing, Specs, and Which to Pick
OpenAI shipped GPT-6 Astra, GPT-6 Sol and Luna, and GPT-6.1 Sol within one month in September 2026. This guide covers the official API pricing for all four models ($10/$50, $2/$10, $0.10/$0.50), the long-prompt pricing rule behind the 1M context window, real cost math, and the code settings to change when migrating from GPT-5.6 — now also covering GPT-6.1 Terra ($1/$5), added on October 26.
AI APIClaude Haiku 5.5 Guide: Tiered API Pricing, Differences from Haiku 4.5, and Migration
Anthropic released Claude Haiku 5.5 on 2026-10-07. For prompts up to 100,000 tokens it costs just $0.10/$0.50 per million tokens, one tenth of Haiku 4.5. But a request whose prompt is over 100,000 tokens is billed at $0.50/$2.50 in full, and the new tokenizer counts the same text as roughly 30% more tokens. This guide covers official specs, both price tiers, cost math that accounts for both traps, and the changes that break or alter behavior when migrating from Haiku 4.5.
AI APIClaude Opus 5.5 and Sonnet 5.5: API Pricing, Migration Changes, and Model Choice
Anthropic released Claude Opus 5.5 (Sept 22) and Sonnet 5.5 (Sept 28) in September 2026. Opus 5.5 is priced at $4/$20 — 20% below Opus 5 — and is now Anthropic's official default for most workloads; Sonnet 5.5 stays at $2/$10. This guide covers official specs, cache pricing, cost math, and the API changes that break code migrating from Opus 5 or Sonnet 5.