Skip to content

Broader API compatibility and model catalog

This major Nexus API release adds model families, expands compatible-client scenarios, and improves the handling of long-running requests.

New models and formats

  • GPT-5.6 Luna, Terra, and Sol are now available.
  • The catalog and Telegram bot include Gemini 3 Flash, Gemini 3.5 Flash, Gemini 3.1 Pro Preview, and Gemini 3.1 Flash Image Preview.
  • OpenAI-compatible, Anthropic-compatible, and Gemini-compatible APIs are available.
  • POST /v1/chat/completions now supports streaming.

Improvements and fixes

Compatibility improved for Codex, Claude Code, VS Code, Zed, JetBrains, OpenCode, bots, and custom applications. Empty responses and format differences in Legacy Completions, Responses Compact, and compact models were fixed.

Long-running and slow requests, pauses between chunks, and early client disconnects are handled more reliably. Errors use one format and include a request ID, while an interrupted or ambiguous response is not treated as fully completed usage.

Most models became approximately 5–15% cheaper. Users who were registered before the release receive an automatic balance bonus.

Compatibility and next steps

Existing keys and base URLs remain valid. Use the API reference to choose a route, check GET /v1/models before configuring a client, and consult the public status page when needed.