Back to rankings

Replit Adds GPT-6 Sol, GPT-6 Luna Fast, and Claude Opus 5.5

ReplitView original changelog

Replit added three new frontier models, OpenAI's GPT-6 Sol and GPT-6 Luna Fast alongside Anthropic's Claude Opus 5.5, to Agent's model lineup just days after both labs shipped them. Developers can now select any of the three directly and tune reasoning effort for supported models, trading speed for depth on a per-task basis. The rollout follows the same-day price cuts both OpenAI and Anthropic made to their new flagship and fast-tier models, giving Replit users access to cheaper, more capable coding models almost immediately after launch.

Key Takeaways

  • Three new frontier models joined Replit Agent within days of their public launch: GPT-6 Sol, GPT-6 Luna Fast, and Claude Opus 5.5.
  • Claude Opus 5.5 leads on coding benchmarks like Terminal-Bench according to independent testing, while running faster and cheaper than the previous Opus.
  • GPT-6 Sol costs about half of OpenAI's flagship Astra model, giving developers a cheaper high-quality option for agentic coding.
  • GPT-6 Luna Fast targets speed over depth, suited to high-volume, low-latency edits rather than complex reasoning chains.
  • New reasoning effort controls let developers trade speed for deeper reasoning on a per-task basis, directly inside Agent.
  • The rollout continues Replit's practice of integrating new frontier models within days of release, keeping Agent's lineup current with the fastest-moving part of the AI industry.

New Frontier Models Land in Agent

Replit updated its model lineup to include three models released within days of each other by OpenAI and Anthropic: GPT-6 Sol, GPT-6 Luna Fast, and Claude Opus 5.5. All three became selectable inside Replit Agent on September 25, just three days after their public launch on September 22, continuing Replit's pattern of adding new frontier models to its picker soon after release.

What Each Model Brings

GPT-6 Sol is OpenAI's general-purpose successor tuned for everyday coding and agentic work, priced at roughly half the rate of OpenAI's top-tier Astra model. GPT-6 Luna Fast is the speed-optimized sibling, built for high-volume, low-latency tasks where quick iteration matters more than maximum reasoning depth. Claude Opus 5.5 is Anthropic's new flagship, which independent benchmarks show performing strongly on coding and agentic tasks such as Terminal-Bench, while also running faster and cheaper than its predecessor.

Configurable Reasoning Effort

Alongside the new models, Replit exposed reasoning effort controls for supported models, letting developers dial thinking time up or down depending on the complexity of the task at hand. Lighter effort settings favor speed for routine edits, while higher settings push the model to reason longer through harder architectural or debugging problems. This gives Replit users the same effort-ladder controls that OpenAI and Anthropic expose in their own consoles, but from directly inside the Agent workspace.

Why It Matters

Both GPT-6 Sol and Luna Fast, as well as Claude Opus 5.5, arrived alongside meaningful price cuts from their respective labs, and Replit passing that access through quickly means developers get the latest reasoning and coding improvements without waiting for a separate rollout cycle. Combined with Replit's existing Auto Mode routing, which picks the best model automatically, users now have a wider and fresher set of models to choose from, whether they want automatic selection or full manual control.

Replit Adds GPT-6 Sol, Luna Fast, Opus 5.5 Models | Yet Another Changelog