AI Coding Tools Roundup: Copilot Fusion, Codex 0.166, Claude Code 2.1.318 and OpenCode 2.2

Executive Summary

• GitHub Copilot Fusion is generally available on all plans. You pick an effort level and Copilot routes each task across models. (changelog)
• Copilot reaches JetBrains Air through the Agent Client Protocol, with local and cloud tasks. (changelog)
• OpenAI Codex CLI 0.165 and 0.166 add GPT-6.1 Terra, a built-in Vertex AI provider and better diagnostics. (Codex changelog)
• Claude Code 2.1.316 to 2.1.318 deprecates fast mode for Opus 4.8 (removal on November 30, 2026) and extends the Claude apps gateway. (CHANGELOG)
• OpenCode 2.2 adds /goal, image generation and a Claude Agent provider, with several breaking changes. (releases)

GitHub Copilot: Fusion and the JetBrains Air integration

Copilot Fusion is generally available

Fusion is a model option, not a model. You select it with an effort level and Copilot decides which models handle each task, combining them when useful. (GitHub changelog)

• Effort levels: Standard for everyday coding, High for harder tasks, Max for the most demanding work.
• Workflows: Single (one model), Cascade (an efficient model drafts and a quality check decides whether to escalate) and Critique (one model drafts, another reviews, the first revises).
• Where it works: VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs.
• Admin control: Fusion only uses models enabled in your enterprise and organization policies, and admins can set it as the default.
• Billing: Each model used is billed at its standard rate. Paid plans get 30% off model costs for a limited time.

GitHub reports that Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max, at about half the average API cost per task. That is a vendor claim, so test it on your own repos before trusting it.

Copilot inside JetBrains Air

Copilot is now an agent in JetBrains Air, using the Agent Client Protocol and your existing Copilot subscription. You can run several sessions in parallel and review changes in context. Cloud tasks run in isolated Docker containers on fresh VMs through Air Teams. Known limits: you can't choose personal or organizational plans per task, Air's governance and analytics don't cover Copilot usage, and admins can't set Copilot spending limits in Air. (GitHub changelog)

Other Copilot entries worth a look

• BYOK and enterprise custom models in the Copilot usage API (Nov 3)
• GitHub Agentic Workflows generally available (Oct 28)
• Agent Merge watches and fixes pull requests in the cloud (Oct 28)
• GPT-6.1 Terra in GitHub Copilot (Oct 26)

OpenAI Codex: GPT-6.1 Terra and a Vertex AI provider

GPT-6.1 Terra is available in Codex on Plus, Pro, Business, Enterprise and Edu plans. OpenAI describes it as a fast, efficient model for professional work. (Codex changelog)

Codex CLI 0.165.0 (October 27)

• Built-in Google Cloud Vertex AI provider, enabled with model_provider = "gcp".
• Repository filter in the session picker and directory grouping in the agent command center.
• Reasoning effort can change mid-conversation by default for supported models.
• Memory v2 is on by default for fresh profiles.

model_provider = "gcp"

Codex CLI 0.166.0 (October 29)

• Keyboard-driven transcript selection with search.
• GPT-6.1 Terra in the bundled, Vertex AI and Amazon Bedrock catalogs.
• New codex queue --resume and codex doctor --network-only.
• Clearer safety-block explanations with safer follow-up prompts, plus fixes for thread resume and fork, app-server races and TUI reconnects.

Claude Code: 2.1.316 to 2.1.318

The Claude Code changelog doesn't date its entries, but these are the three newest versions.

• 2.1.316: Fast mode for Opus 4.8 is deprecated as of October 30, 2026 and will be removed on November 30, 2026. Migrate to Opus 5.5. Also adds CLAUDE_CODE_DISABLE_ULTRATHINK, CLAUDE_CODE_SUBAGENT_EFFORT, pooled group budgets and a layered policy mode for the Claude apps gateway.
• 2.1.317: Warns in /schedule when a new cloud scheduled task gets connectors you didn't request, asks before a web or shell tool call sends your account email over the network, and adds a gateway cost report.
• 2.1.318: Adds assume_role for mantle upstreams on the gateway, shows sign-in expiry in claude auth status, and fixes --continue losing earlier conversation after a malformed tool call.

If you use Opus 4.8 fast mode in scripts, switch before the end of the month.

OpenCode 2.2

OpenCode shipped the 2.2 line between October 27 and 30. (releases)

• v2.2.1: /goal works toward an objective until a reviewer confirms it is met, plus an opt-in image_generate tool.
• v2.2.3: A Claude Agent provider that uses your Claude Code login. TUI plugins now need OpenTUI 0.8.
• v2.2.4 and v2.2.5: Desktop terminals persist per session, experimental artifacts via "artifacts": true, worktree and base-branch shortcuts, and a percentage value for compaction.buffer.

Watch the breaking changes: the OpenCode GitHub Action was removed in 2.2.0, several @opencode/ai providers and entry points were removed in 2.2.1, and integration.use no longer covers MCP servers, skills or plugins.

Also in the ecosystem

• Android Studio Rabbit 2 supports "Bring Your Own Agent", reported to include Claude Agent and Codex. (AI Agents News)
• Runway released Dev plugins for Claude Code, Codex and Cursor. (AI Agents Directory)

Both items come from aggregator sites, so confirm against the vendors' own posts.

What I'd do this week

1. Try Copilot Fusion at High effort on a real task and compare cost and quality with your usual model.
2. Move any Opus 4.8 fast mode usage to Opus 5.5 before November 30.
3. Read the OpenCode 2.2 breaking changes before upgrading, especially if you rely on its GitHub Action or plugins.
4. If you run Codex on Google Cloud, test the new gcp provider.

Sources

• GitHub changelog: Copilot Fusion
• GitHub changelog: Copilot in JetBrains Air
• GitHub Copilot changelog index
• OpenAI Codex changelog
• Claude Code CHANGELOG
• OpenCode releases
• AI Agents News, week of November 3, 2026
• AI Agents News Brief, November 1, 2026