Six New Frontier Models
Samuel Su
on October 10, 2026OpenAI and Anthropic both shipped a new generation, and all six models are now selectable on your agents and through the API. The GPT-6 models are also available in workflow AI nodes. The short version: a new top tier, a Sonnet at the same price as before, and a Haiku at half the price.
What's new
| Model | Context window | Credits / message |
|---|---|---|
| GPT 6 Astra | 1M tokens | 17.5 |
| GPT 6.1 Sol | 1M tokens | 3.5 |
| GPT 6 Luna | 1M tokens | 0.5 |
| Claude Opus 5.5 | 1M tokens | 8 |
| Claude Sonnet 5.5 | 1M tokens | 4 |
| Claude Haiku 5.5 | 1M tokens | 0.5 |
GPT-6 comes in three sizes. Astra is OpenAI's most capable model, for the hardest reasoning, coding and agentic work. Sol is the one most agents should look at: close to Astra on complex questions, at a fraction of the cost. Luna is the volume model, at half a credit per message.
Claude Sonnet 5.5 is Anthropic's best balance of speed and intelligence, at the same four credits per message as Claude 4.6 Sonnet and Sonnet 5, so moving an agent across costs you nothing. Claude Haiku 5.5 is the fastest Claude and, at half a credit per message, a third of the price of Claude 4.5 Haiku. Claude Opus 5.5 is Anthropic's flagship Opus, for complex agentic and enterprise work, at the same eight credits as Claude Opus 5.
Price changes to existing models
One model gets cheaper and a few cost a little more. The changes apply from the next message; you don't need to change anything.
| Model | Before | Now | Newer option for fewer credits |
|---|---|---|---|
| GPT 5.6 Sol | 8 | 6 | — |
| GPT 4o | 2 | 2.5 | GPT 5.4 (2) |
| GPT 5.5 | 8 | 8.5 | GPT 6.1 Sol (3.5) |
| Claude 4.5 Haiku | 1 | 1.5 | Claude Haiku 5.5 (0.5) |
| Claude 4.5 Sonnet | 4 | 4.5 | Claude Sonnet 5.5 (4) |
| Claude 4.5 Opus | 5 | 8 | — |
| Claude 4.6 Opus | 5 | 8 | — |
| Gemini 3.1 Pro | 2.5 | 3.5 | — |
| Gemini 2.5 Pro | 1 | 2 | — |
Why GPT 5.6 Sol is cheaper. OpenAI lowered its price, and we're passing that on.
Why the others cost more. AI providers charge us for every word a model reads and writes. Each message your agent answers carries your instructions, the relevant parts of your knowledge base and the recent conversation, and agents today send more of that than when these prices were set. For these models, a typical message now costs us more than the credits it used. Rather than quietly cut how much your agent can read, we raised each price by the smallest step that covers it. The models, their answers and your settings are unchanged.
If cost matters more than staying on a particular model, the last column lists a newer model that does the same job for fewer credits.
Picking one
If you want a starting point rather than a comparison:
- Most agents → Claude Sonnet 5.5 at four credits, or GPT 6.1 Sol at three and a half. Claude Sonnet 5.5 is a drop-in replacement for any earlier Claude Sonnet, at the same price or less.
- High volume, simple tasks → Claude Haiku 5.5 or GPT 6 Luna at half a credit per message.
- Hardest reasoning → GPT 6 Astra or Claude Opus 5.5. Worth it when the answer matters more than the cost per message; usually not worth it for routine support replies.
Credits per message are shown next to every model in the picker, so you can see what a choice costs before you make it.
Switching
Open your agent, go to the model dropdown, pick one. The change takes effect on the next message — no retraining, no reindexing, nothing to migrate. If it isn't what you wanted, switch back the same way.
The new models are available on plans with advanced-model access.