{"schema":2,"slug":"ai-game-benchmark-2026-07-13","title":"AI Game Benchmark — July 13, 2026","date":"2026-07-13","harness":{"name":"worldbuild-bench","version":"0.1.0","repo_url":"https://github.com/sebnado/worldbuild-bench"},"budgets":{"wall_clock_mins":360,"usd_cap":100},"tracks":[{"slug":"arena-combat","name":"Arena Combat","brief":"Build **Last Stand at the Ruin**, a wave-based 3D arena combat game: the player fights escalating waves of melee enemies inside a closed arena until every wave is cleared or the player dies."},{"slug":"physics-puzzle","name":"Physics Puzzle","brief":"Build **The Marble Works**, a 3D physics puzzle game: the player manipulates objects under gravity to get a ball into a goal zone across a sequence of levels."},{"slug":"racing","name":"Racing","brief":"Build **Sunset Apex**, a 3D circuit racing game: the player races a vehicle against AI opponents over 3 laps on a closed track with checkpoints."}],"results":[{"model_slug":"glm-5.2","model_name":"GLM-5.2","provider":"openai","track_slug":"racing","status":"pass","duration_s":3566,"cost_usd":4.290042,"tokens_in":18779868,"tokens_in_cached":17351517,"tokens_out":267341,"code_lines":5824,"code_files":11,"tool_calls":391,"subagents":11,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-glm-5.2.webp"],"game_id":"3042a88d-9f1a-4d8d-949b-2cd3aef49075","public_game_path":"/worldbuild/games/racing-glm-5.2/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-glm-5.2.webm","capture_ready":true},{"model_slug":"gpt-5.5","model_name":"GPT-5.5","provider":"openai","track_slug":"racing","status":"pass","duration_s":3158,"cost_usd":25.012359,"tokens_in":7609821,"tokens_in_cached":3803648,"tokens_out":135989,"code_lines":4168,"code_files":16,"tool_calls":188,"subagents":7,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-gpt-5.5.webp"],"game_id":"35968e2e-c705-4bd9-82cb-6d659f2db387","public_game_path":"/worldbuild/games/racing-gpt-5.5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-gpt-5.5.webm","capture_ready":true},{"model_slug":"gpt-5.5","model_name":"GPT-5.5","provider":"openai","track_slug":"arena-combat","status":"pass","duration_s":2191,"cost_usd":43.257772,"tokens_in":15731996,"tokens_in_cached":9393664,"tokens_out":228976,"code_lines":8529,"code_files":14,"tool_calls":400,"subagents":14,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-gpt-5.5.webp"],"game_id":"d709f43a-a010-4bc5-85d6-ae5ca6e8e8a6","public_game_path":"/worldbuild/games/arena-combat-gpt-5.5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-gpt-5.5.webm","capture_ready":true},{"model_slug":"gpt-5.5","model_name":"GPT-5.5","provider":"openai","track_slug":"physics-puzzle","status":"pass","duration_s":2419,"cost_usd":23.517063,"tokens_in":18142023,"tokens_in_cached":16636416,"tokens_out":255694,"code_lines":5450,"code_files":17,"tool_calls":352,"subagents":9,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-gpt-5.5.webp"],"game_id":"a80c63de-58cd-4696-99f3-f4020dd27336","public_game_path":"/worldbuild/games/physics-puzzle-gpt-5.5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-gpt-5.5.webm","capture_ready":true},{"model_slug":"glm-5.2","model_name":"GLM-5.2","provider":"openai","track_slug":"arena-combat","status":"pass","duration_s":5758,"cost_usd":6.938141,"tokens_in":26639739,"tokens_in_cached":24380489,"tokens_out":238912,"code_lines":6331,"code_files":11,"tool_calls":411,"subagents":9,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-glm-5.2.webp"],"game_id":"217d872d-21f2-47a3-8570-bb946032e60d","public_game_path":"/worldbuild/games/arena-combat-glm-5.2/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-glm-5.2.webm","capture_ready":true},{"model_slug":"glm-5.2","model_name":"GLM-5.2","provider":"openai","track_slug":"physics-puzzle","status":"pass","duration_s":6547,"cost_usd":7.480741,"tokens_in":26147511,"tokens_in_cached":23028387,"tokens_out":229386,"code_lines":4026,"code_files":11,"tool_calls":352,"subagents":8,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-glm-5.2.webp"],"game_id":"483c4066-1e9d-4f11-a57f-c2df247ba828","public_game_path":"/worldbuild/games/physics-puzzle-glm-5.2/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-glm-5.2.webm","capture_ready":true},{"model_slug":"claude-fable-5","model_name":"Claude Fable 5","provider":"anthropic","track_slug":"arena-combat","status":"pass","duration_s":9711,"cost_usd":118.4306,"tokens_in":61873500,"tokens_in_cached":59700000,"tokens_out":601000,"code_lines":6459,"code_files":10,"tool_calls":147,"subagents":11,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-claude-fable-5.webp"],"game_id":"2cb4c3ae-11cf-40d1-88a8-7c930907e277","public_game_path":"/worldbuild/games/arena-combat-claude-fable-5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-claude-fable-5.webm","capture_ready":true},{"model_slug":"claude-fable-5","model_name":"Claude Fable 5","provider":"anthropic","track_slug":"physics-puzzle","status":"pass","duration_s":32225,"cost_usd":490.6635,"tokens_in":320310100,"tokens_in_cached":313800000,"tokens_out":1800000,"code_lines":7235,"code_files":13,"tool_calls":139,"subagents":15,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-claude-fable-5.webp"],"game_id":"a2106ac3-6a4e-4008-9f90-9d66ee4b99c9","public_game_path":"/worldbuild/games/physics-puzzle-claude-fable-5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-claude-fable-5.webm","capture_ready":true},{"model_slug":"claude-fable-5","model_name":"Claude Fable 5","provider":"anthropic","track_slug":"racing","status":"pass","duration_s":12276,"cost_usd":146.88,"tokens_in":79129200,"tokens_in_cached":76600000,"tokens_out":752700,"code_lines":6767,"code_files":12,"tool_calls":143,"subagents":13,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-claude-fable-5.webp"],"game_id":"409e606a-e20e-4c51-9f50-c23619110bf4","public_game_path":"/worldbuild/games/racing-claude-fable-5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-claude-fable-5.webm","capture_ready":true},{"model_slug":"claude-opus-4.8","model_name":"Claude Opus 4.8","provider":"anthropic","track_slug":"arena-combat","status":"pass","duration_s":5714,"cost_usd":38.2807,"tokens_in":43312200,"tokens_in_cached":42200000,"tokens_out":361800,"code_lines":4639,"code_files":10,"tool_calls":190,"subagents":9,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-claude-opus-4.8.webp"],"game_id":"8c2652a8-b056-4e1a-a8ad-fdc0668ab924","public_game_path":"/worldbuild/games/arena-combat-claude-opus-4.8/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-claude-opus-4.8.webm","capture_ready":true},{"model_slug":"claude-opus-4.8","model_name":"Claude Opus 4.8","provider":"anthropic","track_slug":"physics-puzzle","status":"pass","duration_s":11459,"cost_usd":109.3907,"tokens_in":141098200,"tokens_in_cached":138500000,"tokens_out":751600,"code_lines":5017,"code_files":14,"tool_calls":340,"subagents":11,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-claude-opus-4.8.webp"],"game_id":"9f27c569-ccea-4cc8-8279-a2f576915c5e","public_game_path":"/worldbuild/games/physics-puzzle-claude-opus-4.8/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-claude-opus-4.8.webm","capture_ready":true},{"model_slug":"claude-opus-4.8","model_name":"Claude Opus 4.8","provider":"anthropic","track_slug":"racing","status":"pass","duration_s":9404,"cost_usd":56.4006,"tokens_in":58283500,"tokens_in_cached":56600000,"tokens_out":654600,"code_lines":6781,"code_files":12,"tool_calls":198,"subagents":9,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-claude-opus-4.8.webp"],"game_id":"bf0ea82e-69b1-4f19-a9c8-e60d142dc35a","public_game_path":"/worldbuild/games/racing-claude-opus-4.8/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-claude-opus-4.8.webm","capture_ready":true},{"model_slug":"claude-sonnet-5","model_name":"Claude Sonnet 5","provider":"anthropic","track_slug":"arena-combat","status":"pass","duration_s":5635,"cost_usd":35.6307,"tokens_in":75305100,"tokens_in_cached":73800000,"tokens_out":472800,"code_lines":4947,"code_files":12,"tool_calls":257,"subagents":9,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-claude-sonnet-5.webp"],"game_id":"2c343a69-c9e7-4d6e-847b-955b7d2fc196","public_game_path":"/worldbuild/games/arena-combat-claude-sonnet-5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-claude-sonnet-5.webm","capture_ready":true},{"model_slug":"claude-sonnet-5","model_name":"Claude Sonnet 5","provider":"anthropic","track_slug":"physics-puzzle","status":"pass","duration_s":9546,"cost_usd":90.1507,"tokens_in":226469200,"tokens_in_cached":224000000,"tokens_out":735300,"code_lines":5164,"code_files":15,"tool_calls":685,"subagents":12,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-claude-sonnet-5.webp"],"game_id":"e12261c5-ad0f-4adb-8b9b-374f39449222","public_game_path":"/worldbuild/games/physics-puzzle-claude-sonnet-5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-claude-sonnet-5.webm","capture_ready":true},{"model_slug":"claude-sonnet-5","model_name":"Claude Sonnet 5","provider":"anthropic","track_slug":"racing","status":"pass","duration_s":7475,"cost_usd":41.8906,"tokens_in":82950400,"tokens_in_cached":80800000,"tokens_out":597200,"code_lines":4932,"code_files":12,"tool_calls":169,"subagents":11,"playability":null,"world_coherence":null,"gates":null,"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-claude-sonnet-5.webp"],"game_id":"2caf58d7-13ab-44d4-81f1-d2e61c01ac73","public_game_path":"/worldbuild/games/racing-claude-sonnet-5/index.html","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-claude-sonnet-5.webm","capture_ready":true},{"model_slug":"gpt-5.6-sol","model_name":"GPT-5.6 Sol","provider":"openai","track_slug":"racing","status":"pass","duration_s":5175,"cost_usd":39.673875,"tokens_in":30176583,"tokens_in_cached":27179520,"tokens_out":369960,"code_lines":11973,"code_files":16,"tool_calls":730,"subagents":18,"playability":90,"world_coherence":100,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":true,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":true,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-gpt-5.6-sol.webp"],"game_id":"63e984be-d791-44a0-ad9d-804b181e447e","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-gpt-5.6-sol.webm","capture_ready":true},{"model_slug":"grok-4.5","model_name":"Grok 4.5","provider":"openai","track_slug":"racing","status":"pass","duration_s":1609,"cost_usd":7.13502,"tokens_in":9522078,"tokens_in_cached":8918912,"tokens_out":244872,"code_lines":9242,"code_files":19,"tool_calls":515,"subagents":17,"playability":90,"world_coherence":100,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":true,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":true,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-grok-4.5.webp"],"game_id":"363d76d6-9494-48c1-ad36-b4f252eca422","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-grok-4.5.webm","capture_ready":true},{"model_slug":"gpt-5.6-terra","model_name":"GPT-5.6 Terra","provider":"openai","track_slug":"racing","status":"pass","duration_s":1961,"cost_usd":8.367515,"tokens_in":10520961,"tokens_in_cached":9054208,"tokens_out":162472,"code_lines":3862,"code_files":13,"tool_calls":343,"subagents":12,"playability":90,"world_coherence":100,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":true,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":true,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/racing-gpt-5.6-terra.webp"],"game_id":"5fe843f1-47d2-415c-ac46-2082c3386b3f","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/racing-gpt-5.6-terra.webm","capture_ready":true},{"model_slug":"gpt-5.6-terra","model_name":"GPT-5.6 Terra","provider":"openai","track_slug":"arena-combat","status":"pass","duration_s":2035,"cost_usd":7.291807,"tokens_in":7157865,"tokens_in_cached":5647360,"tokens_out":140247,"code_lines":3953,"code_files":13,"tool_calls":304,"subagents":12,"playability":90,"world_coherence":80,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":true,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":false,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-gpt-5.6-terra.webp"],"game_id":"5eaccb6c-2cbd-4077-943e-1dff42a9b2c3","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-gpt-5.6-terra.webm","capture_ready":true},{"model_slug":"grok-4.5","model_name":"Grok 4.5","provider":"openai","track_slug":"arena-combat","status":"pass","duration_s":1789,"cost_usd":5.121028,"tokens_in":6685031,"tokens_in_cached":6264704,"tokens_out":191337,"code_lines":7292,"code_files":12,"tool_calls":369,"subagents":11,"playability":90,"world_coherence":100,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":true,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":true,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-grok-4.5.webp"],"game_id":"f1d0bb72-503c-40b2-a84d-921e46c9c406","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-grok-4.5.webm","capture_ready":true},{"model_slug":"gpt-5.6-sol","model_name":"GPT-5.6 Sol","provider":"openai","track_slug":"arena-combat","status":"pass","duration_s":3585,"cost_usd":30.859946,"tokens_in":24577270,"tokens_in_cached":22211072,"tokens_out":264114,"code_lines":8489,"code_files":8,"tool_calls":449,"subagents":9,"playability":90,"world_coherence":100,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":true,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":true,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-gpt-5.6-sol.webp"],"game_id":"37873e0c-77e5-4836-801e-b2dc4f5b2274","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/arena-combat-gpt-5.6-sol.webm","capture_ready":true},{"model_slug":"grok-4.5","model_name":"Grok 4.5","provider":"openai","track_slug":"physics-puzzle","status":"pass","duration_s":1698,"cost_usd":6.429278,"tokens_in":8430457,"tokens_in_cached":7893376,"tokens_out":234738,"code_lines":9223,"code_files":13,"tool_calls":396,"subagents":12,"playability":65,"world_coherence":80,"gates":{"loads":true,"no_fatal_errors":false,"webgl_canvas":true,"renders":true,"input_response":false,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":false,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-grok-4.5.webp"],"game_id":"d944e86c-875f-4073-9a91-5a1e90ac9ca1","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-grok-4.5.webm","capture_ready":true},{"model_slug":"gpt-5.6-terra","model_name":"GPT-5.6 Terra","provider":"openai","track_slug":"physics-puzzle","status":"pass","duration_s":2386,"cost_usd":9.536967,"tokens_in":10079803,"tokens_in_cached":8065536,"tokens_out":165661,"code_lines":4496,"code_files":12,"tool_calls":318,"subagents":12,"playability":75,"world_coherence":100,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":false,"stable_fps":false,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":true,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-gpt-5.6-terra.webp"],"game_id":"1f4f75a6-a702-484e-acc0-5d5613e9bd3e","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-gpt-5.6-terra.webm","capture_ready":true},{"model_slug":"gpt-5.6-sol","model_name":"GPT-5.6 Sol","provider":"openai","track_slug":"physics-puzzle","status":"pass","duration_s":4308,"cost_usd":37.553359,"tokens_in":28618013,"tokens_in_cached":25561088,"tokens_out":316273,"code_lines":9767,"code_files":14,"tool_calls":712,"subagents":18,"playability":100,"world_coherence":100,"gates":{"loads":true,"no_fatal_errors":true,"webgl_canvas":true,"renders":true,"input_response":true,"stable_fps":true,"restart_works":true,"bench_contract":true,"wc_position_sane":true,"wc_camera_sane":true,"wc_entities_query":true,"wc_state_evolves":true,"wc_reset_returns":true},"screenshots":["https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-gpt-5.6-sol.webp"],"game_id":"6e82c553-9f43-4090-8030-cd875d7605fe","public_capture_path":"https://gamemedia.sandscape.app/worldbuild/2026-07/physics-puzzle-gpt-5.6-sol.webm","capture_ready":true}],"notes":["The generated Playability and World Coherence gate scores are intentionally omitted from the public report because this round showed that they do not reliably reflect human-obvious quality failures.","The quality signal for this round is the WorldBuild Arena: anonymous pairwise human evaluation of which game people would rather keep playing."],"stage":"pilot","arena_ratings":null}