One prompt, three games. Claude Fable 5.2, Claude Opus 5.5 and GPT-6.1 Sol on high effort.
Claude Fable 5.2 and Claude Opus 5.5 in Claude Code and GPT-6.1 Sol in Codex CLI each built a first-person voxel sandbox game from the same prompt and no picture: a generated block world to walk through, mine and build in. One run each on high effort, with no human edits.
This is not a CmdBench benchmark run: no checks, scores or verdicts.
The three games
One run per model. Each game is served exactly as the agent wrote it, with no human edits, and opens full screen in this tab.
Claude Fable 5.2
Tessel Isle: an island of 192 × 192 blocks, drawn with Three.js and its own shaders, with nine blocks to build with.
Click the game to look around; each game shows its own controls. Made for a mouse and keyboard. Opens full screen in this tab.
The same scripted minute of play, in four frames
0:02Before the first click
0:19Mining
0:33Building
0:59The sky, later
In the same scripted minute of play, all three games took the mouse and keys: each one locks the mouse to look around, walks, jumps, mines the block its highlight marks, picks a block by number key and by mouse wheel, places blocks, and changes its sky as the minute passes.
One difference between the runs: Codex CLI can't be limited to file tools, so the GPT-6.1 Sol run could also run commands in a sandbox (its own folder, no network, no browser). Its log shows it ran none; it only wrote files.
Costs are estimates at API list prices. For Claude Fable 5.2 and Claude Opus 5.5 they are Claude Code's own. Codex CLI reports no cost, so GPT-6.1 Sol's is computed from the run's own token counts at OpenAI's list prices of 09.11.2026.
One run per model: exploratory, not a ranking.
The prompt
Build a first-person voxel sandbox game that runs in a desktop browser: a world made of blocks that the player walks through, mines and builds in.
Claude Fable 5.2 (claude-fable-5-2), Claude Opus 5.5 (claude-opus-5-5), GPT-6.1 Sol (gpt-6.1-sol)
Agents
Claude Code 2.1.322 (Claude Fable 5.2 and Claude Opus 5.5); Codex CLI 0.167.1 (GPT-6.1 Sol)
Effort
high, one run per model
Tools
Claude Code had file tools only (Read, Write, Edit, Glob, Grep): no browser, no shell, no web. Codex CLI can't be limited to file tools, so the GPT-6.1 Sol run could also run commands in a sandbox: its own folder only, no network, no browser. Its log shows it ran none; it only wrote files. No model saw its game running.
Time limit
60 minutes per run
Three.js
0.187.0, in the build folder: the only library the prompt allowed
Human edits
None. Each game is served exactly as the agent wrote it.