OpenAI · Released September 22, 2026

Use GPT-6 Sol in AimiChat

GPT-6 Sol is OpenAI's mid-tier GPT-6 model for demanding reasoning, coding and agent work. In AimiChat it powers the Premium quality, available from the Aimi Ultra plan.

Chat with GPT-6 Sol
Plan
Aimi Ultra · $50/month
Quality setting
Premium
Context window
1.05M tokens (922K input)
API price
$2 input / $10 output per 1M tokens

What is GPT-6 Sol?

OpenAI released GPT-6 Sol on September 22, 2026. It sits below the flagship GPT-6 Astra and is the upgrade path for people who used GPT-5.6 Sol: roughly half the price, with OpenAI describing it as the model for serious analysis and coding that does not need Astra's maximum intelligence.

Do not confuse it with GPT-5.6 Sol, which is an older, separate model still available through the API. GPT-6 Sol has a knowledge cutoff of April 20, 2026 and the same 1.05M-token context window as the rest of OpenAI's current models.

GPT-6 Sol specs and pricing

Made byOpenAI
ReleasedSeptember 22, 2026
API model IDgpt-6-sol
Context window1.05M tokens (922K input)
Max output128,000 tokens
Knowledge cutoffApril 20, 2026
InputsText and images
API price$2 input / $10 output per 1M tokens
In AimiChatPremium quality · Aimi Ultra ($50/month)

API prices are what developers pay OpenAI directly. In AimiChat you pay a flat plan price instead and do not need your own API key. Limits per message and per day depend on your plan.

What GPT-6 Sol is good at

  • Multi-step reasoning. Analysis with several moving parts: planning, trade-off decisions, finance and data questions.
  • Agentic coding. Longer coding tasks that need a plan, several edits and a check that the result works.
  • Math and science. Worked solutions where each step needs to be shown and checked.
  • Research synthesis. Combining many sources into one structured answer when paired with web search.

When to use a different model

  • Cheap, quick questions. For everyday chat, GPT-6 Luna is much faster and is included on the Free plan.
  • Polished prose. For writing that needs a natural voice, many people prefer Claude Sonnet 5 or Claude Opus 5.5.

GPT-6 Sol benchmarks

BenchmarkScoreReported by
DeepSWE v1.1 (max effort)68.8%OpenAI launch post, as reported by VentureBeat and Codersera
OSWorld 2.0 offline (xhigh effort)60.5%OpenAI launch post, as reported by VentureBeat and Codersera

These are the maker's own published results. Different companies run different test setups, so scores are only comparable within the same table from the same source.

Prompts to try with GPT-6 Sol

  1. I have to choose between these two job offers. Build a decision table, tell me what evidence would change the answer, then recommend one.
  2. Plan the refactor of this module into smaller files, list the risks, then write the first step.
  3. Solve this problem step by step and check your final answer a second way.

How to use GPT-6 Sol in AimiChat

  1. Open AimiChat and sign in on the Aimi Ultra plan or higher (see plans).
  2. Choose the Premium quality in the model menu next to the message box.
  3. Send your message. You can switch quality in the same chat at any time, so you can start with a faster model and move up when a question needs more depth.

Compare GPT-6 Sol

See every model in AimiChat side by side.

GPT-6 Sol FAQ

Which AimiChat plan includes GPT-6 Sol?

GPT-6 Sol powers the Premium quality, included from the Aimi Ultra plan ($50/month). Aimi Genius includes it too.

Is GPT-6 Sol the same as GPT-5.6 Sol?

No. GPT-6 Sol (September 2026) is a newer, cheaper mid-tier model. GPT-5.6 Sol was the flagship of the older GPT-5.6 family.

What is GPT-6 Sol's knowledge cutoff?

April 20, 2026, according to OpenAI's model reference. Use web search in AimiChat for anything more recent.

How does GPT-6 Sol compare to Claude Opus 5.5?

Opus 5.5 is Anthropic's flagship and costs twice as much through the API. Sol is the value pick for reasoning and coding. See the full comparison for a task-by-task breakdown.

Sources

Sources checked · September 24, 2026

Specifications and prices come from the model makers and API listings below. Benchmark figures are reported by the company that made the model and are not independent tests.