Claude Sonnet 5.5 and GPT-6 Join the Catalog
October 1, 2026 · InfraScout Team
The balanced tier gets new models from both providers. Claude Sonnet 5.5 replaces Sonnet 5 on Anthropic connections, and OpenAI's GPT-6 family replaces the GPT-5.6 tiers. Both are selectable now in the model catalog. Each one starts at a medium reasoning effort, and the new default for OpenAI connections costs half as much as the old one.
Claude Sonnet 5.5
Sonnet 5.5 takes Sonnet 5's place in the picker and is now the default for new Anthropic connections. It has one million tokens of context and 128k of output. It costs exactly what Sonnet 5 did, on every line of the bill:
| Claude Sonnet 5 | Claude Sonnet 5.5 | |
|---|---|---|
| Input, per million tokens | $2.00 | $2.00 |
| Output, per million tokens | $10.00 | $10.00 |
| Cached input, per million tokens | $0.20 | $0.20 |
The model costs the same per call, but how many calls it makes depends on effort. At high effort, Sonnet 5.5 makes noticeably more tool calls per question than Sonnet 5 did, so each answer costs more even though each call costs the same. Anthropic recommends medium effort for multi-step tool use, and that is what Sonnet 5.5 runs at when a chat is set to Auto.
The GPT-6 family
On OpenAI connections, three GPT-6 models replace the GPT-5.6 tiers in the picker. All three have a one-million-token context window:
| Model | Role | Input | Cached input | Output |
|---|---|---|---|---|
| GPT-6.1 Sol | Balanced, the new default | $2.00 | $0.10 | $10.00 |
| GPT-6 Astra | Most capable | $10.00 | $1.00 | $50.00 |
| GPT-6 Luna | Fast and low-cost | $0.10 | $0.01 | $0.50 |
Prices are per million tokens. GPT-5 nano stays in the picker as the lowest-cost option.
GPT-6.1 Sol is now the default for new OpenAI connections, at $2 input and $10 output. That is half of what GPT-5.6 Sol, the previous default, costs — even after OpenAI cut GPT-5.6 Sol from $5 / $30 to $4 / $20. Cost estimates for connections still on GPT-5.6 Sol follow that cut.
GPT-6 models run at medium reasoning effort on Auto, which is OpenAI's own default for the family. The GPT-5 family stays at high. The same mapping applies when the model is served from an Azure AI Foundry deployment.
Long-context pricing on OpenAI
OpenAI charges more for requests with over 272K input tokens: double the input rate and one and a half times the output rate. InfraScout's cost estimates do not reflect this surcharge yet. Very long OpenAI conversations can cost more than the estimate shows.
Auto, and when to change it
Auto is a starting point, not a limit. If a question needs a deeper investigation, you can choose a higher effort level for that chat in the effort picker. If it is a quick lookup, you can choose a lower one. The Chat guide shows where to find the picker. For how the tiers compare across providers, see Choosing a Model.
Your existing connections
Adding a model never moves an existing connection onto it. A connection on Sonnet 5 keeps working at the same price. Under Admin › AI Integrations › AI Providers, it now shows an Outdated model badge naming Sonnet 5.5 as its replacement. Whether to switch is up to you.
OpenAI connections get the same treatment. GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.5, GPT-5.4 and GPT-5 point to GPT-6.1 Sol. GPT-5.6 Luna, GPT-5.4 mini, GPT-5.4 nano and GPT-5 mini point to GPT-6 Luna. To switch, an administrator edits the connection's model under AI Provider Connections.
Older Claude models no longer fail
Some connections still ran on older Claude models, such as Claude 3.x or Sonnet and Opus 4.0 through 4.5. These connections could fail on every turn, because they received settings their model does not support: effort needs Opus 4.5 or Sonnet 4.6 and later, and adaptive thinking needs Opus or Sonnet 4.6 and later. InfraScout now sends each setting only to models that accept it, so these connections run at their model's own defaults and answer normally — no change needed on your side. Moving each of them to a current model is still worthwhile, as its Outdated model badge suggests.
For the Opus tier, see Claude Opus 5.5 Replaces Opus 5 and Opus 4.8.
Questions or feedback? Reach us at info@infrascout.cloud.