GPT-6.2 API: Sol vs Astra
Compare GPT-6.2 Sol and Astra on TTAPI: everyday and complex workflows, Chat Completions versus Responses, token prices, and cache-write billing.
What you will take away
- Choose a model for everyday or complex work
- Compare all five token billing dimensions
- Keep Chat Completions and Responses requests aligned
Sol is OpenAI's efficient everyday model; Astra targets the hardest projects. Compare finished-task quality and total token usage. TTAPI applies its existing OpenAI 20% discount to both models, including the documented cache-write rates.
Two models, different priorities
TTAPI added gpt-6.2-sol and gpt-6.2-astra on November 18. OpenAI presents Sol for everyday work and Astra for demanding projects. Start with the task's acceptance criteria, then compare the finished result, latency, and billed usage.
OpenAI highlights improvements in coding, professional documents, visual work, and token efficiency. These are provider claims, rather than a TTAPI benchmark. A useful trial gives both models the same source material and review criteria.
GPT-6.2 Sol
gpt-6.2-solAn efficient default for recurring coding, writing, and agent tasks.
- Scoped implementation and review
- Repeated document and analysis tasks
- Lower base token prices
GPT-6.2 Astra
gpt-6.2-astraA candidate for difficult engineering, research, and professional deliverables.
- Complex software investigations
- Multi-step professional work
- Higher base rates than Sol
Both use TTAPI's documented OpenAI routes and the existing 20% channel discount.
Token prices and cache writes
The table shows TTAPI base prices in USD per million tokens, after its 20% OpenAI discount. Cache writes and cached input are separate billing dimensions. The detail pages show the official base rates alongside TTAPI prices.
OpenAI's model references describe higher rates above 272K input tokens: input and cache rates double, and output costs 1.5 times the short-context rate for the full request. Processing tiers and tools can add charges. Confirm TTAPI's current billing rules before estimating those workflows.
| Billing item / 1M tokens | GPT-6.2 Sol | GPT-6.2 Astra |
|---|---|---|
| Input | $1.6 | $6.4 |
| Cache write · 30m | $2 | $8 |
| Cache write · 12h | $4 | $16 |
| Cached input | $0.08 | $0.32 |
| Output | $8 | $32 |
A Chat Completions request
For the TTAPI chat route, send messages and the exact model ID, with the optional stream boolean. This example is a text review request; tool calling requires Responses for GPT-6.2.
curl --request POST \
--url 'https://api.ttapi.io/v1/chat/completions' \
--header "Authorization: Bearer $TTAPI_KEY" \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6.2-sol",
"messages": [
{
"role": "user",
"content": "Review this migration plan and identify three risks with an acceptance check for each."
}
],
"stream": false
}'A Responses request
The Responses route uses input instead of messages. The example selects medium reasoning and leaves tools out. Add supported tools only after checking both the model and gateway references.
TTAPI's generic schema includes options that GPT-6.2 does not accept. Omit temperature, top_p, top_logprobs, and message.output_text.logprobs. OpenAI also changed the model's cache-retention configuration; confirm TTAPI support before supplying cache controls from a different schema.
curl --request POST \
--url 'https://api.ttapi.io/v1/responses' \
--header "Authorization: Bearer $TTAPI_KEY" \
--header 'Content-Type: application/json' \
--data '{
"model": "gpt-6.2-astra",
"input": "Review this migration plan and identify three risks with an acceptance check for each.",
"reasoning": {
"effort": "medium"
},
"stream": false
}'