On November 17, 2026, OpenAI released GPT-6.2 Astra (gpt-6.2-astra) and GPT-6.2 Sol (gpt-6.2-sol) across ChatGPT Work, Codex, and the API.
Short answer: GPT-6.2 Astra costs $8 input / $40 output per 1M tokens, exactly 20% less than GPT-6 Astra ($10 / $50). GPT-6.2 Sol stays at $2 / $10, the same as GPT-6.1 Sol, with a newer knowledge cutoff. All figures below come from OpenAI's official pricing page and model pages, retrieved November 18, 2026.
GPT-6.2 vs. the previous versions
Item | GPT-6.2 Astra | GPT-6 Astra | GPT-6.2 Sol | GPT-6.1 Sol |
|---|---|---|---|---|
Input (1M tokens) | $8 | $10 | $2 | $2 |
Output (1M tokens) | $40 | $50 | $10 | $10 |
Cached input | $0.40 | $1.00 | $0.10 | $0.10 |
Cache write (30 min) | $10 | $12.50 | $2.50 | $2.50 |
Context | 1,050,000 for all (max input 922,000, max output 128,000) | |||
Knowledge cutoff | May 18, 2026 | Apr 30, 2026 | Jun 1, 2026 | Apr 30, 2026 |
The biggest change for Astra users is cached input: $1.00 → $0.40. Long system prompts and reused documents get much cheaper. Above 272K input tokens, the whole request is billed at 2x input and cache rates and 1.5x output, as before.
Processing tiers: no Flex row, Ultrafast at 3x for now
Tier | GPT-6.2 Astra | GPT-6.2 Sol | Note |
|---|---|---|---|
Standard | $8 / $40 | $2 / $10 | Base |
Batch | $4 / $20 | $1 / $5 | 50% off |
Fast | $16 / $80 | $4 / $20 | 2x |
Ultrafast | $24 / $120 | $6 / $30 | 3x through Feb 28, 2027; 6x from Mar 1, 2027 (PT) |
Flex | No row on the pricing page | GPT-6 Astra and 6.1 Sol still have one | |
If you run jobs on Flex today, moving them to 6.2 as-is doubles the rate. The alternative OpenAI added the same day is the background completion window in the Responses API (public beta): 30% off for a 2-hour window and 40% off for an 8-hour window, for GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra, and GPT-6 Luna.
Monthly cost: three example cases
Monthly input / output | GPT-6 Astra | GPT-6.2 Astra | GPT-6.2 Sol |
|---|---|---|---|
Internal chatbot (10M / 2M) | $200 | $160 | $40 |
Coding agent (50M / 10M) | $1,000 | $800 | $200 |
Same, via 8-hour window | n/a | $480 | $120 |
Short context, Standard, no caching. Token counts can change too: OpenAI says GPT-6.2 Astra writes more concisely by default, which lowers output cost.
Check this before migrating: when reasoning items get dropped
OpenAI's GPT-6.2 guide says that starting with 6.2, the API drops reasoning items when it detects a mid-conversation prefix edit, a model downgrade, or a change of end user or organization. Quality can drop and token use can rise. Two common traps:
- Injecting a timestamp or counter into the system prompt each turn. Move variable values to the latest user message.
- Planning with Astra and executing with Sol inside the same conversation. Split the conversations.
Also, reasoning.effort: none is rejected. Replace none or minimal with low. Tool calling requires the Responses API.
Not confirmed yet
- Benchmark numbers (we could not retrieve the announcement body).
- Which ChatGPT plans get 6.2 and when.
- Retirement dates for GPT-6 Astra and GPT-6.1 Sol (none listed on the deprecations page as of Nov 18).
Same price as Claude now: see GPT-6.2 vs. Claude Fable 5.2 pricing.
FAQ
How much does GPT-6.2 Astra cost?
$8 per 1M input tokens and $40 per 1M output tokens on Standard, 20% less than GPT-6 Astra. Cached input fell from $1.00 to $0.40.
Is GPT-6.2 Sol more expensive than GPT-6.1 Sol?
No. Both are $2 input, $10 output and $0.10 cached input. The knowledge cutoff moved from April 30 to June 1, 2026.
Can I use GPT-6.2 with Flex processing?
As of November 18, 2026 the pricing page has no Flex row for GPT-6.2, and the migration guide lists Standard, Fast and Batch. The new background completion window (30% or 40% off) is the closest alternative.