GPT-6.2 Sol API
NewLLMThe GPT-6.2 Sol API is a newer snapshot of OpenAI's cost-oriented flagship line, billed at 9% of OpenAI's list: 1M context, vision and tool use, with every meter following the list's step above 272K.
- Input · <=272K$0.18 USDOfficial price
$2 USD - Output · <=272K$0.9 USDOfficial price
$10 USD - Input · >272K$0.36 USDOfficial price
$4 USD - Output · >272K$1.35 USDOfficial price
$15 USD - Savings91%
- Context1M
- View tool pricing and billing rules
What are we making today?
Bring an idea, a question, or a first draft.
01
What this model is for
The GPT-6.2 Sol API is a newer snapshot of OpenAI's cost-oriented flagship line, billed at 9% of OpenAI's list: 1M context, vision and tool use, with every meter following the list's step above 272K.
GPT-6.2 Sol is a newer snapshot of the cheaper OpenAI line that the catalog carries as GPT-6.1 Sol, and it sits below GPT-6.2 Astra in the same family: same 1M window and the same vision and tool-use support, now on a cheaper card. The gateway bills this line at roughly a tenth of OpenAI's list price on every meter, and above 272K it follows the list's own step rather than a flat doubling. That makes it the cheapest Sol snapshot so far — well below the GPT-6.1 Sol entry on input, and further below it on output. The catalog quotes that step per band: one quote up to 272K and the list's step above.
In our 18 November 2026 runs a merge-intervals coding prompt returned 166, 168 and 147 output tokens over three calls, for a median of 166 — a little shorter than the same prompt on GPT-6.2 Astra in the same session. JSON mode returned a parseable object, tool calling produced a valid get_weather call for Lisbon, and image input correctly described a supplied cover on the first attempt. That mix is what the Sol line is for: predictable, cheap agent steps rather than the most expensive reasoning available.
Best-fit coding workloads
High-volume agent loops
Use it for the many small steps inside an agent — classify, extract, decide, call a tool — and reserve the Astra snapshot for the steps that actually need reasoning depth. Practical tip: instrument cost per completed task, not cost per call, before assuming the cheaper line wins.
Batch extraction and tagging
Run thousands of documents through one fixed schema with JSON mode and keep going while the output parses. Practical tip: fail loudly on the first unparseable record instead of writing nulls, because a silent schema drift is far more expensive than a retry.
Everyday coding assistance
Wire it into Codex or Cursor for edits, test fixes and small refactors where the diff is short. Practical tip: keep the context to the file plus its direct callers; the window is large, but a small context makes the review smaller too.
02
Hands-on notes
Checked on 2026-11-18 through the OmniaKey API with one standard coding prompt (median of 3 runs) and three capability probes.
- Output tokens for our standard coding prompt, reasoning included
- 166
- Cost of that output at current prices
- $0.00015
- JSON mode
- Passed
- Tool calling
- Passed
- Image input
- Passed
On 18 November 2026 three runs of a merge-intervals coding prompt returned 166, 168 and 147 output tokens, a median of 166.
The same session passed a JSON-mode request, produced a get_weather tool call for Lisbon, and described a supplied cover image correctly on the first attempt.
03
Technical specification
Catalog facts and integration details verified for this route.
- Model ID
- gpt-6.2-sol
- Context
- 1M
- Inputs and tools
- Vision · Tool use
- API endpoint
- /v1/chat/completions
- Cache hit / 1M
- $0.009 USD
- Coding clients
- Codex · Cursor · Cline · Aider
Limits to plan around
- The cheaper card is a price change, not a spec change: it answers like the previous Sol snapshot but costs less, so any cost comparison built on the old Sol rates needs re-running.
- It is not the flagship: on the hardest reasoning tasks the Astra snapshot answered the same prompt with a little more elaboration, and we did not run a scored comparison.
- Nothing is aliased. Clients must name the new id, and a setup that mixes Sol and Astra snapshots needs its own way to attribute spend per model.
04
Compare nearby models
Compare workload fit and current listed token rates before choosing a route.
| Model | Best for | Context | Input | Output |
|---|---|---|---|---|
| GPT-6.2 SolThis page gpt-6.2-sol | The newest Sol snapshot, billed at roughly a tenth of OpenAI's list price with the same long-context step; the default for volume work now. | 1M | $0.18 USD Official price $2 USD | $0.9 USD Official price $10 USD |
| GPT-6.1 Sol gpt-6.1-sol | The previous Sol snapshot on the older, higher card; keep it as the baseline your evaluations compare against. | 1M | $0.225 USD Official price $2 USD | $1.464 USD Official price $10 USD |
| GPT-6.2 Astra gpt-6.2-astra | The flagship sibling from the same launch: same window, higher rate card, and its own step above 272K. Route only the steps that need the extra reasoning here. | 1M | $0.72 USD Official price $8 USD | $3.6 USD Official price $40 USD |
05
SDK quickstarts
Use one base URL and model ID from cURL, Python, or TypeScript.
curl https://api.omniakey.com/v1/chat/completions \
-H "Authorization: Bearer $OMNIAKEY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-6.2-sol",
"messages": [
{"role": "user", "content": "Review this pull request"}
],
"stream": true
}'06
Estimate token cost
Enter expected input and output tokens to compare current listed rates.
- OmniaKey estimate
- $0.036
- Official-rate estimate
- $0.40
- Estimated savings
- $0.364
Estimate covers input and output tokens only. Cache usage, retries, tools, and client-side token accounting can change the final charge.
07
Related coding guides
Deeper selection, setup, and troubleshooting guidance for this model family.
08
Sources and verification
Model facts and links were checked on 2026-11-18. Provider documentation remains the authority for upstream behavior.
Frequently asked questions
How much does GPT-6.2 Sol cost on OmniaKey?
Rates vary with input context length: <=272K: Input $0.18 USD, Output $0.9 USD, Cache hit / 1M $0.009 USD; >272K: Input $0.36 USD, Output $1.35 USD, Cache hit / 1M $0.018 USD.
What is the context window for GPT-6.2 Sol?
The listed context window is 1M. Actual usable limits can also depend on the client and upstream request format.
How does GPT-6.2 Sol differ from GPT-6.2 Astra?
They share the launch date, the 1M window, vision and tool use, but not the rate card: Sol keeps the cheaper Sol rates and halved cached input, while Astra carries the flagship rates and a different discount in the catalog. Use Sol for volume and Astra for the steps where a wrong answer costs more than the tokens.
Does GPT-6.2 Sol cost more than GPT-6.1 Sol?
No — it is cheaper. GPT-6.2 Sol runs on its own lower card, billed at roughly a tenth of OpenAI's list price, while GPT-6.1 Sol keeps the older card, so input, output and cached input all cost less for the same kind of work; the price table on this page shows both cards per context band.
Should I switch everything to GPT-6.2 Sol?
Only after replaying the evaluation set. It answered our merge-intervals prompt slightly more tersely than GPT-6.2 Astra and passed JSON, tool calling and image input, but a snapshot can lose a case your tests cover and we cannot see. Keep a GPT-6.1 Sol baseline so a regression is visible immediately.
How was GPT-6.2 Sol checked before it was listed?
We ran a merge-intervals coding prompt three times on 18 November 2026 and recorded 166, 168 and 147 output tokens, then tested JSON mode, tool calling and image input on the same route. The counts are in the measurement table and the guide states what we did not test rather than rounding it away.