TL;DR: We have added support for Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna. You can select them directly in chat or assign them to your dedicated agents.

Frontier model releases usually come with long waiting periods. This week, Anthropic and OpenAI broke that cadence by dropping major updates right on the heels of their previous launches.

Anthropic rolled out Claude Opus 5.5 just sixty days after releasing Opus 5 in July. Initial developer reaction across technical communities has focused on its efficiency gains: Opus 5.5 beats the much pricier Claude Fable 5.1 across agentic coding and knowledge benchmarks, while cutting token pricing by 20% and reducing typical workload costs by roughly 40%. The biggest unlock for builders is step reduction, with the model completing multi-turn agent loops in noticeably fewer tool calls.

At the same time, OpenAI moved quickly to flesh out its GPT-6 family. Just three weeks after introducing its top-tier flagship GPT-6 Astra, OpenAI launched GPT-6 Sol and GPT-6 Luna. The public response has been immediate relief on pricing: while Astra pushed frontier intelligence, developers wanted practical workhorses for real production pipelines. Sol delivers near-flagship software engineering capabilities at half the cost, while Luna provides fast, low-cost execution with tunable reasoning depth.

Both providers are shifting the frontier from pure raw scale to operational efficiency. Here is how each new model fits into your Starchild workspace.

1. Claude Opus 5.5: Flagship Agentic Coding and Fewer Steps

Claude Opus 5.5 is Anthropic's new flagship for complex codebases and long-running autonomous agents.

• What it delivers: Higher accuracy on agentic coding benchmarks than Fable 5.1, paired with lower API rates and aggressive prompt caching discounts. Because it plans more concisely, it resolves tasks in fewer steps, lowering both turnaround time and total token consumption.
• Best workflow fit: Multi-file architectural refactoring, end-to-end debugging sessions, deep analytical research, and autonomous agents executing chained tool calls.
• How to use it: Choose Claude Opus 5.5 in chat when tackling intricate software problems, or assign it to dedicated agents that manage long-horizon coding projects.

2. GPT-6 Sol: High-End Software Engineering at Half the Cost

GPT-6 Sol serves as OpenAI's mainstream flagship alternative for serious engineering work.

• What it delivers: Near-flagship reasoning and code generation without the premium cost of GPT-6 Astra. It cuts token prices by roughly 50% compared to earlier top-tier tiers, making deep reasoning viable for daily engineering.
• Best workflow fit: Day-to-day full-stack development, code reviews, technical documentation, and professional workflows where lightweight models lack depth but full flagships are cost-prohibitive.
• How to use it: Use Sol as your daily driver for technical tasks in chat, or configure it as the core reasoning engine for development agents.

3. GPT-6 Luna: Fast, Budget Execution with Tunable Reasoning

GPT-6 Luna is OpenAI's high-speed, budget-conscious model built for high-throughput pipelines.

• What it delivers: Minimal latency, high throughput, and rock-bottom token pricing, alongside adjustable reasoning levels so you can dial in extra depth when a task requires it.
• Best workflow fit: Bulk classification, scheduled ingestion scripts, rapid triage routines, conversational bots, and large-scale data transformation runs.
• How to use it: Point background monitor jobs, scheduled scrapers, and high-frequency sub-agents at Luna to keep operational spend predictable.

Summary: Matching Models to Tasks

You can start using Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna immediately from the model picker in chat or configure them directly inside your agent settings.