{"slug": "claude-code-in-unity-we-benchmarked-it-on-a-game", "title": "Claude Code in Unity: We Benchmarked It on a Game", "summary": "In a benchmark of AI coding agents building a Flappy Bird game in Unity, Claude Opus 5 scored 10/10 and shipped a working game with a 306-line PlayMode test suite, while Claude Sonnet 5 scored 5/10 because it placed the ground at y = −250, making the game unplayable. DeepSeek v4 Flash, a free model, delivered a full game for $0.61, and DeepSeek v4 Pro matched the 9/10 score but cost $11.96 due to zero cache reads. The benchmark, run by Ziva in the Unity editor beta, used three prompts with a 20-minute cap per model.", "body_md": "# Claude Code in Unity: We Benchmarked It on a Game\n\nZiva runs docked in the Unity editor now, in beta, so we reran\n[our Godot benchmark](/blogs/gpt-5-6-benchmark-godot) there: build a playable Flappy Bird as\nan agent, three prompts, 20-minute cap each. Claude ran on a **Pro/Max subscription**.\n\n| Model | Tool calls | Wall time | Build cost | Rubric | Stars |\n|---|---|---|---|---|---|\n| DeepSeek v4 Flash (free) | 158 | 45.6 min (2 capped) | $0.61 | 9/10 | 3/5 |\n| DeepSeek v4 Pro | 413 | 42.9 min (1 capped) | $11.96* | 9/10 | 3/5 |\nClaude Sonnet 5 (BYO) | 384 | 28.2 min | $7.34* | 5/10 | 2/5 |\nClaude Opus 5 (BYO) | 213 | 33.6 min | $14.73* | 10/10 | 4/5 |\n\n* List-price equivalents, not billed: Claude ran on a subscription and this account has a 100%-off promotion on Pro. Flash’s 61 cents is real spend.\n\n**Sonnet 5** wrote the cleanest code and a game you cannot play. The scripts are correct;\nthe scene is not. The camera sees y −5 to +5; Sonnet put the Ground at **y = −250**, so the\nbird dies before the first pipe arrives.\n\n**Opus 5** shipped a working game and, unprompted, a 306-line PlayMode suite — twelve tests,\nrun sixteen times, ending at thirteen passing. One failure (“an in-flight keypress must not\nskip the overlay”) is why it added a restart delay.\n\n**DeepSeek v4 Flash**, the free model, shipped a full game for 61 cents — built entirely in\ncode — but capped on two turns. Same prompts cost **$0.18 in Godot**.\n\n**DeepSeek v4 Pro** matched that 9/10 in 413 calls, but made **zero cache reads** to Flash’s\n10.4 million — most of the twenty-fold cost gap.\n\nUnity went from **12 tools to 21** this release, adding `run_tests`\n\n, `find_gameobjects`\n\nand\n`manage_asset`\n\n. Two of four models used `run_tests`\n\nunasked.\n\nCaveats: n=1 per model, one task type, wall times not comparable across engines. One\nUnity-beta defect cost three runs — a changed-API edit raises a modal dialog that blocks the\neditor, and `-accept-apiupdate`\n\ndoesn’t stop it. The beta is in\n[Ziva](/blogs/best-ai-tools-for-godot-2026).", "url": "https://wpnews.pro/news/claude-code-in-unity-we-benchmarked-it-on-a-game", "canonical_source": "https://ziva.sh/blogs/claude-code-unity-benchmark", "published_at": "2026-08-04 00:00:00+00:00", "updated_at": "2026-08-05 19:56:14.085735+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "ai-agents", "developer-tools"], "entities": ["Ziva", "Unity", "Claude Opus 5", "Claude Sonnet 5", "DeepSeek v4 Flash", "DeepSeek v4 Pro", "Godot"], "alternates": {"html": "https://wpnews.pro/news/claude-code-in-unity-we-benchmarked-it-on-a-game", "markdown": "https://wpnews.pro/news/claude-code-in-unity-we-benchmarked-it-on-a-game.md", "text": "https://wpnews.pro/news/claude-code-in-unity-we-benchmarked-it-on-a-game.txt", "jsonld": "https://wpnews.pro/news/claude-code-in-unity-we-benchmarked-it-on-a-game.jsonld"}}