Guide

AI Coding Tools 2026: What Changed This Month (October Tracker)

What changed in Cursor, GitHub Copilot, Claude Code, Codex and Devin Desktop in October 2026: model retirements, deadlines, new features and current plan prices.

AI Coding Tools 2026: What Changed This Month (October Tracker)

Short answer: the October 2026 change most likely to break your setup is a model retirement, not a feature. GitHub Copilot drops Claude Opus 5, Claude Opus 4.8 and GPT-5.6 Terra on November 16, and Codex already retired GPT-5.5 for ChatGPT sign-ins on October 14. Copilot Pro still costs $10 a month, and Claude Code shipped Haiku 5.5 at $0.10 per million input tokens.

This tracker logs what five widely used coding tools changed between late September and October 18, 2026, read on each vendor's own changelog. Each entry carries its date, what it means for your workflow, and whether you have to act. Feature launches you can ignore are marked as such, because most of a changelog is noise for a working team.

Toolradar data: among the 40 AI coding tools with enough traffic to rank on Toolradar's AI coding statistics (snapshot of October 16, 2026), Cursor drew 439 compare-page views in 30 days, ahead of GitHub Copilot at 267 and Claude Code at 210. Those are the tools buyers are weighing against each other right now, so they lead this log.

What changed this month at a glance

Tool Biggest change (date) Do you need to act? Entry paid plan
GitHub Copilot Three models deprecated on November 16 (announced October 15) Yes, before November 16 Pro, $10/mo
OpenAI Codex GPT-5.5 retired for ChatGPT sign-ins (October 14) Yes, update saved model settings ChatGPT Plus, $20/mo
Claude Code Claude Haiku 5.5 added as the default Haiku (October 7) Only if you pin model IDs Claude Pro, $17/mo billed annually
Cursor Origin App Marketplace opens to public submissions (October 13) No Plus, $8/mo (Grok models only)
Devin Desktop (Windsurf) Enterprise extension marketplace URL and newer MCP support (October 12) Admins only Not covered in this log

Prices in this table were checked on October 18, 2026, on Cursor's pricing page, GitHub Copilot plans, Claude pricing and ChatGPT pricing. The full price history per tool sits on our AI coding assistant pricing page.

Which deadlines can break your setup?

Four dates matter more than every feature launch combined, because each one changes what a saved config, script or CI job is allowed to call. Put them in the team calendar now.

November 16, 2026, GitHub Copilot. Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra leave every Copilot surface, from chat and agent mode to code completions, per the deprecation notice. GitHub names Claude Opus 5.5 as the replacement for both Opus models, and GPT-6.1 Sol or GPT-6 Luna for Terra. One exception helps solo buyers: individual subscribers on annual plans keep Opus 4.8.

October 14, 2026, Codex. GPT-5.5 retired from Codex for users signed in with ChatGPT, per the Codex changelog. OpenAI asks you to update workspace defaults, managed configurations, custom agents, scheduled tasks and scripts that still select gpt-5.5. API-key sessions are not affected, so a CI job that authenticates with a key keeps running.

November 21, 2026, Codex credits. The Codex pricing page says GPT-5.6 Sol's promotional credit pricing runs at least through that date. If a team budget assumes the promo rate, recheck the credit table after it.

December 14, 2026, Claude Design. Claude Code 2.1.302 added a notice to /design-sync and /design-login that the standalone Claude Design site closes by that date, per the Claude Code releases. Anyone syncing design files through those commands should plan the move before mid-December.

GitHub Copilot: models in, models out, and code review billing

Copilot had the busiest October of the five, and most entries concern models rather than features. That pattern matters because Copilot now bills model use through GitHub AI Credits, so a model swap changes cost as well as output.

  • October 7, Claude Haiku 5.5 arrives. GitHub's announcement pitches it for subagents, quick edits and terminal tasks, on Pro, Pro+, Max, Business and Enterprise, billed at provider list pricing under usage-based billing.
  • October 13, Gemini 4 Argon arrives. The release note limits it to Pro+, Max, Business and Enterprise, so a Pro seat does not get Google's newest model.
  • October 7, local sandboxing is generally available. Copilot CLI, the Copilot app and VS Code Agent Host sessions can now restrict file, network and credential access for agent-run commands, at no extra cost, per the sandboxing note. Teams that let agents run shell commands should turn it on.
  • October 8, code review billing moves. Organization owners can bill Copilot code reviews to the organization instead of each member's quota, per the code review update. The default still bills the member, and a member whose quota is exhausted gets a failed review.
  • October 14, HydraFusion Max. The multi-model orchestration option now draws on Claude Opus 5.5, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, and stays in research preview, per GitHub's note.

What it costs now: Pro comes with $15 of AI Credits each month: base credits equal to its price plus a $5 flex allotment that GitHub says may change. Pro+ at $39 carries $70 of credits, and Max at $100 carries $200. Code completions do not use credits on paid plans, so the credit pool only drains on chat, agents, CLI and code review. The Copilot pricing breakdown covers what each tier unlocks.

Claude Code: two new default models and idle compaction

Claude Code shipped near-daily releases, and two of them changed which model you get by default. If your scripts pass a model ID explicitly, nothing changes; if they rely on the default alias, the model under them moved.

  • September 28, Claude Sonnet 5.5 (claude-sonnet-5-5) became the default Sonnet on the Anthropic API in version 2.1.284, with 1M context at $2 input and $10 output per million tokens.
  • October 7, Claude Haiku 5.5 (claude-haiku-5-5) became the default Haiku in 2.1.293, with 1M context at $0.10 input and $0.50 output per million tokens, rising to $0.50 and $2.50 for prompts over 100K tokens.
  • October 14, idle compaction. Version 2.1.299 compacts a conversation over 200K tokens that sits idle for about 54 minutes, before its one-hour prompt cache expires. That saves a full reprocess on your next message, and "idleCompaction": false turns it off.
  • October 1, Claude Mods. Version 2.1.287 lets plugins modify deeper behavior, which is the change to watch if your team writes its own tooling around the CLI.

What it costs now: Claude Code is included from Claude Pro, at $17 a month billed annually ($200 up front) or $20 monthly, and Max starts at $100 a month for 5x or 20x Pro usage. The Claude Code pricing guide works through when Max pays off, and how to use Claude Code covers setup.

OpenAI Codex: GPT-6.1 Sol becomes the default path

Codex spent the month moving users from GPT-5.5 to GPT-6.1 Sol, which landed on September 29 and which OpenAI describes as near-Astra performance at a lower cost than Astra. The retirement on October 14 is the forcing function, so expect any old pinned config to fail rather than fall back quietly.

  • October 8, GPT-6.1 Sol Ultrafast. Faster token generation, limited to the Pro $500 tier and eligible Enterprise and Edu plans, and off by default for Enterprise until a workspace owner enables it.
  • October 9, composer predictions (beta). The desktop app suggests your next message after each response; Tab accepts it as an editable draft. It is limited to adults on personal ChatGPT Pro plans in supported regions.
  • October 15, Codex CLI 0.163.0. A searchable /config, steering a running turn by sending a follow-up, and parallel read-only tool calls.

What it costs now: ChatGPT Plus is $20 a month with expanded Codex usage, Go is $8, and Pro starts at $100 with longer Codex sessions. For the head to head, see Codex vs Claude Code.

Cursor: Origin grows an app store

Cursor's October changes build out Origin, its code-hosting product, rather than the editor. That is useful context if you are still pricing Cursor as an editor subscription, because the plans now sell a wider bundle.

  • October 6, remote control for local agents. The Cursor iOS app can see and reply to agents running on your own computer, on by default for everyone except Enterprise organizations. The computer has to stay on and online.
  • October 13, Origin App Marketplace. Origin integrates with over 20 apps across deployments, CI, code review, security, monitoring and project management, and the marketplace now takes public submissions. Our Origin launch coverage explains the product.
  • October 16, cloud agent canvases and the Environments API (beta). Cloud agent environments now support up to 300 repositories and load large environments 5x faster, per Cursor's changelog.

What it costs now: Plus is $8 a month but only offers Grok models, Premium is $30 with 4x Plus usage, and Super and Ultra are $100 and $200. Third-party models are billed separately by usage on every plan, so a developer who wants Claude or GPT inside Cursor pays on top of the subscription. The Cursor pricing page has the full table, and the SpaceX deal explains the Grok bundle.

Devin Desktop (Windsurf): admin controls and MCP

Windsurf's editor now ships as Devin Desktop, and its October release is aimed at administrators rather than individual developers. Version 3.10.67 on October 12 lets Enterprise admins set an extension marketplace URL so the editor installs extensions from that source only, and adds support for MCP servers on the 2026-07-28 protocol revision, per the Devin Desktop changelog.

The bigger September change was billing: version 3.10.48 lets you sign in with ChatGPT and bill GPT model requests to an eligible Plus or Pro plan. If you already pay for ChatGPT, that removes a second model bill for GPT use inside the editor. Plan prices are not part of this log; the Windsurf review covers the editor's plans.

How to read this tracker every month

Read the "do you need to act" column first, then the deadlines section, and skip feature launches that do not touch your stack. Feature notes are worth a look only when a tool you pay for moved a capability between plans, as Gemini 4 Argon did by skipping Copilot Pro.

Developers who want these changes in their inbox the day they ship can take the daily Devshot note. Subscribe here.

How we built this log

We read each vendor's own changelog and release notes between September 25 and October 18, 2026: Cursor's changelog, the GitHub Copilot changelog, the Claude Code releases, the Codex changelog and the Devin Desktop changelog. Prices come from each vendor's pricing page on October 18. Performance claims ("near-Astra", "faster") are the vendor's words, not our measurements.

The order follows what developers act on: deadlines first, then changes by tool. Louis Corneloup, founder of Dupple and Toolradar, edits this log, and Devshot's editorial rules are on the about page. For a ranked view rather than a change log, see the best AI coding assistants and the best AI coding agents.

FAQ

What changed in AI coding tools in October 2026?

The main changes were model retirements and new default models. GitHub Copilot announced that Claude Opus 5, Claude Opus 4.8 and GPT-5.6 Terra leave Copilot on November 16, Codex retired GPT-5.5 for ChatGPT sign-ins on October 14, and Claude Code made Haiku 5.5 its default Haiku on October 7. Cursor opened its Origin App Marketplace to public submissions on October 13.

Which Copilot models are being deprecated in November 2026?

Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra, on November 16, across chat, inline edits, agent mode and code completions. GitHub suggests Claude Opus 5.5 for the two Opus models and GPT-6.1 Sol or GPT-6 Luna for Terra. Individual subscribers on annual plans keep Claude Opus 4.8.

What should I do about the GPT-5.5 retirement in Codex?

Switch saved settings, managed configurations, custom agents, scheduled tasks and scripts from gpt-5.5 to gpt-6.1-sol where your plan offers it. The retirement only applies to Codex sessions signed in with ChatGPT; sessions authenticated with an OpenAI API key are not affected.

Which AI coding tool is cheapest to start with in October 2026?

Cursor Plus is the lowest paid sticker at $8 a month, but it only offers Grok models, and other models are billed separately. GitHub Copilot Pro at $10 a month includes $15 of AI Credits, with code completions not drawing on them. ChatGPT's free plan includes limited Codex access, and Copilot Free supports the CLI and agent mode with limited usage.

Does Claude Code cost extra on top of a Claude plan?

No. Claude Code is included from the Pro plan, which is $17 a month billed annually or $20 monthly, and on every Max tier.

How often is this tracker updated?

Each month, after reading the five vendors' changelogs again. Every entry keeps the date the vendor published it, so an older change stays traceable after a newer one replaces it.

Cite this: Devshot, "AI Coding Tools 2026: What Changed This Month (October Tracker)", October 2026.

Some offers on this page may be paid placements or contain affiliate links.

Devshot: the daily dev & AI brief

Free daily newsletter, read in 5 minutes.

Subscribe free