OpenAI GPT-6 Sol and GPT-6 Luna Are Now Available

Shortly after GPT-6 Astra, OpenAI has released two more GPT-6 models: GPT-6 Sol and GPT-6 Luna. Both are now in the AI Data Stream model picker. GPT-6 Sol is the most capable OpenAI option at standard pricing, one tier below the premium GPT-6 Astra, and GPT-6 Luna is the new low-cost choice for quick questions.

Pricing

GPT-6 Sol (per million tokens):

• Input: $2
• Cached input: $0.20
• Cache writes: $2.50
• Output: $10

GPT-6 Luna (per million tokens):

• Input: $0.10
• Cached input: $0.01
• Cache writes: $0.125
• Output: $0.50

Both models have a context window of just over one million tokens. Compared with the models they replace, and with GPT-5.6 Sol, which stays in the picker:

• GPT-6 Sol vs GPT-5.6 Terra ($2 / $12): the same input price and cheaper output, for a stronger model.
• GPT-6 Luna vs GPT-5.6 Luna ($0.20 / $1.20): less than half the price.
• GPT-6 Sol vs GPT-5.6 Sol: half the price. GPT-5.6 Sol is currently on promotional pricing of $4 / $20, which OpenAI says runs at least through November 21, 2026; its standard price is $5 / $30.

Two details matter for the way this app works:

• Caching softens the cost of long conversations. Every follow-up question re-reads the conversation so far, and OpenAI bills that repeated prefix at the cached rate - a tenth of the normal input price. The first time a prefix is cached it is billed as a cache write at 1.25x the input rate. Caching is automatic; there is nothing to switch on.
• Very large requests cost more. OpenAI bills requests with more than 272K input tokens at 2x the input rate and 1.5x the output rate. Normal analytics conversations sit far below that, but it is worth knowing before pasting in very large documents.

As always, AI Data Stream is bring-your-own-key: you are billed by OpenAI directly at their published rates.

What’s New

OpenAI trained Sol and Luna with similar methods to GPT-6 Astra, bringing Astra’s improvements to cheaper models. The points that matter most for analytics:

• Fewer factual mistakes. On OpenAI’s internal factuality evaluation, built from real conversations where users flagged an error, GPT-6 Sol makes about half as many mistakes as GPT-5.6 Sol. GPT-6 Luna at higher effort levels matches GPT-5.6 Sol on the same evaluation.
• Better at multi-step business work. On AutomationBench, Zapier’s benchmark of business workflows that span several apps, GPT-6 Luna at high effort scores 5.4 percentage points above its predecessor at 58% lower cost per task.
• Clearer answers. OpenAI says both models inherit Astra’s communication style: less jargon, fewer low-value details and slightly shorter answers.
• Changing effort keeps the cache. According to OpenAI, raising or lowering the reasoning effort mid-conversation no longer resets prompt caching. You can switch effort levels between questions without paying full price for the conversation so far.

See OpenAI’s announcement for the full details.

GPT-5.6 Terra and Luna Have Been Retired

GPT-5.6 Terra and GPT-5.6 Luna are no longer in the picker. If either was your default model, nothing breaks and no action is needed:

• GPT-5.6 Terra → GPT-6 Sol
• GPT-5.6 Luna → GPT-6 Luna

Both switches lower what you pay per token. GPT-5.6 Sol stays in the picker, labelled as the previous generation, so you can compare it with GPT-6 Sol on your own data before moving over. It costs twice as much per token, so it is worth making that comparison soon.

New Default OpenAI Model

GPT-6 Sol is now the default when you add a new OpenAI provider, replacing GPT-4o. If you have already chosen a default model, it stays as it is (unless it was Terra or 5.6 Luna, as above).

Reasoning Effort Control

GPT-6 Sol and GPT-6 Luna offer every effort level in the chat picker:

• Low - Faster responses, lower cost
• Medium - Balanced speed and depth
• High - Thorough reasoning for complex analysis
• Extra high - Deeper reasoning for difficult, multi-step questions
• Max - Maximum reasoning for the hardest questions

Reasoning tokens are billed as output, so higher levels cost more per answer. On OpenAI’s figures, Luna at a higher effort level can match GPT-5.6 Sol’s factual accuracy at a small fraction of the cost, so it is worth trying before reaching for a bigger model.

The Full OpenAI Lineup

• GPT-6 Sol - Advanced: the most capable option at standard pricing (default)
• GPT-5.6 Sol - Previous generation
• GPT-6 Astra - Most powerful, premium; for the hardest analysis
• GPT-6 Luna - Fast and very low cost
• GPT-5 Mini - Fast, budget-friendly
• GPT-4o - Legacy model without reasoning
• GPT-4o Mini - Fastest, budget-friendly

Switching Models

1. Go to Settings > AI Providers
2. Click your OpenAI provider
3. Select your preferred model from the dropdown
4. Save

Or simply select it from the model picker when starting a new conversation.

Questions?

If you have any questions, contact us or check our AI providers documentation.