Claude Opus 5.5, GPT-6 Sol and Luna: what's new, now in iolys
See what's new in Claude Opus 5.5, GPT-6 Sol and Luna. Use them natively in iolys through Claude Code, Codex, Kiro or OpenRouter.
Claude Opus 5.5, GPT-6 Sol and GPT-6 Luna arrived on September 22, 2026. Opus improves coding, speed and writing. Sol and Luna bring lower API prices for complex work and smaller tasks. Anthropic announcement, OpenAI release notes.
Opus 5.5: better coding and less waiting
Anthropic reports four improvements over Opus 5:
- Stronger coding: better results on large code changes and performance investigations help with demanding refactoring work.
- Faster output: text generation is more than 30% faster, so you wait less for responses.
- Clearer writing: responses put key information first, making the work easier to review.
- Better respect for your instructions: fewer actions outside the assigned task help keep changes within scope. Opus also better resists malicious instructions found in external content.
API input and output prices are 20% lower, and cache reads cost 60% less. With reduced token use, Anthropic estimates 40% lower API costs for typical work at default settings. Your savings depend on the task. Opus 5.5 performance, speed and pricing.
GPT-6 Sol: lower API costs for complex work
GPT-6 Sol is built for demanding coding tasks that need planning, tools and checks. Its standard API input and output prices are 50% lower than GPT-5.6 Sol's, helping you spend less on complex work. GPT-6 Sol specifications, GPT-5.6 Sol pricing.
Try Sol for a feature implementation, a difficult bug or a detailed code review. OpenAI also recommends it for research and deeper analysis of code or documents. OpenAI model selection guide.
GPT-6 Luna: lower API costs for routine tasks
GPT-6 Luna suits clear, repeatable tasks. Compared with GPT-5.6 Luna, standard API input prices fall 50% and output prices fall about 58%, giving you lower costs for routine work. GPT-6 Luna specifications, GPT-5.6 Luna pricing.
Based on OpenAI's guidance, try it for small edits, extracting information or writing a first draft. In a C# project, start with a documentation update or a simple code change with clear requirements. OpenAI model selection guide.
API prices at a glance
Standard Anthropic and OpenAI API prices in USD per million tokens, checked September 23, 2026:
| Model | Input | Cached input | Output |
|---|---|---|---|
| Claude Opus 5.5 | $4.00 · ↓ −20% | $0.20 · ↓ −60% | $20.00 · ↓ −20% |
| GPT-6 Sol | $2.00 · ↓ −50% | $0.20 · ↓ −50% | $10.00 · ↓ −50% |
| GPT-6 Luna | $0.10 · ↓ −50% | $0.01 · ↓ −50% | $0.50 · ↓ −58.3% |
↓ Lower price compared with each model's previous version: Opus 5, GPT-5.6 Sol and GPT-5.6 Luna, respectively. Cached input prices are for cache reads.
For Sol and Luna, these rates apply to prompts with up to 272K input tokens. Longer prompts cost more. Cache writes and optional tools have separate charges. These are API prices. Your subscription price and allowance depend on your provider's plan. Anthropic pricing announcement, Sol pricing details, Luna pricing details.
All three are natively available in iolys
All three models are natively available in iolys for Visual Studio 2026. Choose your provider and stay in Visual Studio:
- Claude Opus 5.5: use your Claude subscription through Claude Code, or your Kiro subscription.
- GPT-6 Sol and GPT-6 Luna: use your ChatGPT subscription through Codex, or your Kiro subscription.
- All three models: use OpenRouter with your API key and account credits.
For Claude Code, Codex or Kiro, connect your account in Manage Providers. For OpenRouter, add your API key in Manage Models. Then select a model for your conversation. Access and usage limits follow your provider account and plan.
Try the models on your C# and .NET projects, with your solution context, tools and permissions available in iolys.
Share your experience on Discord
Tried Opus 5.5, Sol or Luna in iolys? Join us on Discord to share your feedback, ask questions or tell us which model works best for your projects.