TL;DR

OpenCode is the best free, model-agnostic option — 100% open source (MIT), works with 75+ model providers, and you only pay for the AI models you plug in. Claude Code has the strongest raw model quality, now running Fable 5 and Opus 4.8 with computer use. Cursor is still the most polished all-in-one IDE, but it's now a SpaceX subsidiary after a $60 billion all-stock acquisition — worth watching for pricing or roadmap changes ahead.

AI coding tools had their biggest month of 2026 in June. SpaceX bought Cursor's parent company for $60 billion days after its own record-breaking IPO. OpenCode quietly became the most-adopted open-source coding agent ever built, crossing 7.5 million monthly developers. And Claude Code gained access to Anthropic's most capable models plus computer-use browser control. If you're choosing (or re-evaluating) an AI coding tool right now, here's what actually changed and which one fits your workflow.

What Changed This Month

  • SpaceX agreed to acquire Cursor's parent company, Anysphere, for $60 billion in an all-stock deal announced June 16, 2026 — one of the largest acquisitions of a venture-backed startup ever
  • The deal came just days after SpaceX's own Nasdaq debut, the biggest IPO in history at $75 billion raised
  • OpenCode crossed 160,000+ GitHub stars and 7.5 million monthly active developers, becoming the top pick in several dev-tool power rankings for the first time since Cursor's own 3.0 rebuild
  • Claude Code added computer use (opening apps and navigating browsers directly from the terminal) and now runs both Opus 4.8 and Fable 5, giving it the highest coding-arena Elo of any tool on this list

At a Glance

OpenCodeCursorClaude Code
PriceFree (open source)$20/moUsage-based via API/plan
Cost in practice$2–$64/mo in model API usage$20/mo flatDepends on model + usage
License / ownershipMIT, community-builtOwned by SpaceX (pending close)Owned by Anthropic
Models supported75+ providers (Claude, GPT, Gemini, DeepSeek, local)Claude, GPT-4o and othersOpus 4.8, Fable 5
Runs whereTerminal, IDE, or desktopVS Code forkTerminal, with computer use
Adoption7.5M monthly developers~50,000 enterprise clientsAnthropic subscriber base
Best atModel flexibility, self-hostingPolished all-in-one IDERaw model quality, agentic depth

OpenCode — Best Free, Model-Agnostic Option

OpenCode is built by the team behind SST (now Anomaly) and is 100% open source under the MIT license — no paid tier, no premium plan, no usage cap on the tool itself. You bring your own API keys for whichever models you want (Claude, GPT, Gemini, DeepSeek) or run everything locally via Ollama at zero marginal cost. Real-world spend typically lands between $2 and $64 a month depending on how much you code and which models you pick.

What sets it apart technically is LSP integration that feeds compiler and linter diagnostics straight back to the model — something no other major coding agent does — plus background subagents, a Scout agent for external research, and true air-gapped deployment for regulated industries. For developers who want maximum control over cost and model choice, OpenCode is currently unmatched.

Cursor — Most Polished IDE, Now a SpaceX Subsidiary

Cursor remains the most refined standalone AI IDE — full codebase awareness, a Composer mode for multi-file edits, and a workflow that feels like a natural evolution of VS Code rather than a bolt-on chatbot. It reached roughly $4 billion in annualized revenue in under four years, with about $2.6 billion from enterprise customers spanning two-thirds of the Fortune 500.

The wrinkle: SpaceX's $60 billion acquisition of parent company Anysphere is expected to close in Q3 2026, making Cursor a wholly owned subsidiary of Elon Musk's rocket company. At roughly 15x revenue, it's one of the largest multiples ever paid for an AI software company. Nothing has changed for users yet, but a new owner this large — with priorities well outside developer tooling — is a real variable for anyone building long-term workflows on Cursor.

Claude Code — Best Model Quality

Claude Code is Anthropic's terminal-based coding agent, and it now gives you a choice between Opus 4.8 and Fable 5 — the same frontier model that completed a 50-million-line Ruby migration for Stripe in a single day. With computer use layered in, Claude Code can open apps and navigate a browser directly from the terminal, closing a real gap with IDE-based tools. It currently posts the highest WebDev Arena Elo (1653) of any tool in this comparison, reflecting raw model capability rather than tooling polish.

Which One Should You Use?

  • Want full control over cost and models, or need self-hosted/air-gapped deployment → OpenCode
  • Want the most polished all-in-one IDE experience and don't mind a flat monthly fee → Cursor
  • Want the single highest-quality model for hard agentic or long-horizon coding tasks → Claude Code
  • Working in a regulated or security-sensitive environment → OpenCode's air-gapped mode
  • Uncertain about Cursor's roadmap post-acquisition → worth trialling OpenCode or Claude Code as a hedge
Verdict

There's no single winner — these three now serve different priorities. OpenCode is the best value and the most flexible, Cursor is still the most polished IDE (with a genuine ownership question mark hanging over it), and Claude Code has the strongest models underneath. If you have to pick one today with no other context, start with OpenCode: it's free to try, model-agnostic, and lets you swap in Claude Code's own models if you want that quality without leaving the terminal.