Minimal model updates.
Google updated its Gemini 3 Pro model to version 3.1 with improved high-level agent capabilities. Anthropic updated its mid-tier Claude Sonnet 4.5 to 4.6. These are gradually being added to all major AI coding tools. The Qwen3.5-Plus model was added to Qwen Code.

https://blog.kilo.ai/p/grok-code-fast-optimized
https://kilo.ai/landing/grok-code-fast-1-optimized
xAI ended the free giveaway of Grok Code Fast 1 via Kilo as of January 20, but added (and then temporarily removed) an optimized free version.

Copilot Subscription in Zed
https://github.blog/changelog/2026-02-19-github-copilot-support-in-zed-generally-available/
GitHub officially enabled the use of Copilot Pro, Pro+, Business, or Enterprise subscriptions in Zed through a partnership. Authentication happens directly via the GitHub Copilot account—no additional license or separate API key is required.

Blocking from Anthropic
https://code.claude.com/docs/en/legal-and-compliance
Since January–February 2026, users have encountered bans on their Pro/Max subscriptions when used in non-Anthropic tools.

This is now officially documented in the Legal & Compliance section: OAuth tokens from Free, Pro, and Max plans are intended exclusively for the official Claude Code and Claude.ai. Using these tokens in any third-party tools, editors, or services is prohibited. Accounts will be blocked without warning.

https://news.ycombinator.com/item?id=47069299
Approximately 80% of comments are critical of Anthropic. The decision is seen as classic "enshittification" and a lock-in attempt to artificially force all users into their own Claude Code, which has recently become less convenient (especially the decision to hide the model's reasoning/thinking) compared to OpenCode, Cursor, Codex, Aider, etc. This will likely accelerate the transition to alternatives.

#newllmmodel #kilo #githubcopilot #zed #claudecode