{
  "benchmark": "Terminal-Bench 2 difficult subset operational probe",
  "run_at": "2026-07-08",
  "dataset": "terminal-bench/terminal-bench-2",
  "agent": "terminus-2",
  "host": "Forge/MS-01 via Harbor",
  "tasks": [
    "cancel-async-tasks",
    "write-compressor",
    "feal-differential-cryptanalysis"
  ],
  "oracle_smoke": "passed for all three selected tasks",
  "important_caveat": "Small difficult subset; not leaderboard-equivalent Terminal-Bench 2.1 score. Grok 4.5 FEAL should be retried once before using as final capability claim because it hit APIError. The Grok 4.3 FEAL wrapper was interrupted with exit code 130 after Harbor had been running that task for 30m53s; publish it as interrupted/incomplete, not a scored capability failure.",
  "summary": {
    "GPT-5.5": {
      "passes": 3,
      "fails": 0,
      "exceptions": 0,
      "interrupted_incomplete": 0,
      "attempted_tasks": 3,
      "model_slug": "openai/gpt-5.5",
      "known_cost_usd": 0.601997
    },
    "Grok 4.5": {
      "passes": 1,
      "fails": 1,
      "exceptions": 1,
      "interrupted_incomplete": 0,
      "attempted_tasks": 3,
      "model_slug": "xai/grok-4.5",
      "known_cost_usd": 0
    },
    "Grok 4.3": {
      "passes": 1,
      "fails": 0,
      "exceptions": 1,
      "interrupted_incomplete": 1,
      "attempted_tasks": 3,
      "model_slug": "xai/grok-4.3",
      "known_cost_usd": 0.67763
    }
  },
  "rows": [
    {
      "model": "GPT-5.5",
      "model_slug": "openai/gpt-5.5",
      "task": "cancel-async-tasks",
      "job": "gpt55-t2-cancel-async-tasks",
      "status": "pass",
      "reward": 1.0,
      "exceptions": {},
      "started_at": "2026-07-08T14:40:22.250553",
      "finished_at": "2026-07-08T14:41:37.448516",
      "updated_at": "2026-07-08T14:41:37.448516",
      "duration_s": 75.2,
      "cost_usd": 0.06160000000000001,
      "input_tokens": 2696,
      "output_tokens": 1604,
      "cache_tokens": 0,
      "completed_trials": 1,
      "errored_trials": 0,
      "running_trials": 0
    },
    {
      "model": "GPT-5.5",
      "model_slug": "openai/gpt-5.5",
      "task": "write-compressor",
      "job": "gpt55-t2-write-compressor",
      "status": "pass",
      "reward": 1.0,
      "exceptions": {},
      "started_at": "2026-07-08T14:47:06.750559",
      "finished_at": "2026-07-08T14:49:14.117677",
      "updated_at": "2026-07-08T14:49:14.117677",
      "duration_s": 127.4,
      "cost_usd": 0.21222000000000002,
      "input_tokens": 14910,
      "output_tokens": 4589,
      "cache_tokens": 0,
      "completed_trials": 1,
      "errored_trials": 0,
      "running_trials": 0
    },
    {
      "model": "GPT-5.5",
      "model_slug": "openai/gpt-5.5",
      "task": "feal-differential-cryptanalysis",
      "job": "gpt55-t2-feal-differential-cryptanalysis",
      "status": "pass",
      "reward": 1.0,
      "exceptions": {},
      "started_at": "2026-07-08T14:49:16.375618",
      "finished_at": "2026-07-08T14:52:42.747981",
      "updated_at": "2026-07-08T14:52:42.747981",
      "duration_s": 206.4,
      "cost_usd": 0.328177,
      "input_tokens": 40931,
      "output_tokens": 8303,
      "cache_tokens": 27904,
      "completed_trials": 1,
      "errored_trials": 0,
      "running_trials": 0
    },
    {
      "model": "Grok 4.5",
      "model_slug": "xai/grok-4.5",
      "task": "cancel-async-tasks",
      "job": "grok45-t2-cancel-async-tasks",
      "status": "fail",
      "reward": 0.0,
      "exceptions": {},
      "started_at": "2026-07-08T14:42:58.196333",
      "finished_at": "2026-07-08T14:45:00.181009",
      "updated_at": "2026-07-08T14:45:00.181009",
      "duration_s": 122.0,
      "cost_usd": null,
      "input_tokens": 12406,
      "output_tokens": 6464,
      "cache_tokens": 7936,
      "completed_trials": 1,
      "errored_trials": 0,
      "running_trials": 0
    },
    {
      "model": "Grok 4.5",
      "model_slug": "xai/grok-4.5",
      "task": "write-compressor",
      "job": "grok45-t2-write-compressor",
      "status": "pass",
      "reward": 1.0,
      "exceptions": {},
      "started_at": "2026-07-08T14:52:44.655937",
      "finished_at": "2026-07-08T15:00:01.427624",
      "updated_at": "2026-07-08T15:00:01.427624",
      "duration_s": 436.8,
      "cost_usd": null,
      "input_tokens": 69737,
      "output_tokens": 28080,
      "cache_tokens": 50816,
      "completed_trials": 1,
      "errored_trials": 0,
      "running_trials": 0
    },
    {
      "model": "Grok 4.5",
      "model_slug": "xai/grok-4.5",
      "task": "feal-differential-cryptanalysis",
      "job": "grok45-t2-feal-differential-cryptanalysis",
      "status": "exception",
      "reward": 0.0,
      "exceptions": {
        "APIError": [
          "feal-differential-cryptanalysis__T3RGNDj"
        ]
      },
      "started_at": "2026-07-08T15:00:03.344035",
      "finished_at": "2026-07-08T15:02:01.151852",
      "updated_at": "2026-07-08T15:02:01.151852",
      "duration_s": 117.8,
      "cost_usd": null,
      "input_tokens": 3044,
      "output_tokens": 4808,
      "cache_tokens": 896,
      "completed_trials": 1,
      "errored_trials": 1,
      "running_trials": 0
    },
    {
      "model": "Grok 4.3",
      "model_slug": "xai/grok-4.3",
      "task": "cancel-async-tasks",
      "job": "grok43-t2-cancel-async-tasks",
      "status": "pass",
      "reward": 1.0,
      "exceptions": {},
      "started_at": "2026-07-08T14:41:46.288709",
      "finished_at": "2026-07-08T14:42:50.926405",
      "updated_at": "2026-07-08T14:42:50.926405",
      "duration_s": 64.6,
      "cost_usd": 0.007267900000000001,
      "input_tokens": 6164,
      "output_tokens": 981,
      "cache_tokens": 2752,
      "completed_trials": 1,
      "errored_trials": 0,
      "running_trials": 0
    },
    {
      "model": "Grok 4.3",
      "model_slug": "xai/grok-4.3",
      "task": "write-compressor",
      "job": "grok43-t2-write-compressor",
      "status": "exception",
      "reward": 0.0,
      "exceptions": {
        "AgentTimeoutError": [
          "write-compressor__QGaWZJL"
        ]
      },
      "started_at": "2026-07-08T15:02:03.251531",
      "finished_at": "2026-07-08T15:17:42.645966",
      "updated_at": "2026-07-08T15:17:42.645966",
      "duration_s": 939.4,
      "cost_usd": 0.47044080000000005,
      "input_tokens": 664534,
      "output_tokens": 44365,
      "cache_tokens": 448704,
      "completed_trials": 1,
      "errored_trials": 1,
      "running_trials": 0
    },
    {
      "model": "Grok 4.3",
      "model_slug": "xai/grok-4.3",
      "task": "feal-differential-cryptanalysis",
      "job": "grok43-t2-feal-differential-cryptanalysis",
      "status": "interrupted_incomplete",
      "reward": null,
      "exceptions": {
        "Interrupted": [
          "wrapper_exit_130_after_30m53s"
        ]
      },
      "started_at": "2026-07-08T15:17:44.608987",
      "finished_at": null,
      "updated_at": "2026-07-08T20:48:37.799353Z",
      "duration_s": 1853.0,
      "cost_usd": 0.19992105000000004,
      "input_tokens": 395329,
      "output_tokens": 17134,
      "cache_tokens": 321024,
      "completed_trials": 1,
      "errored_trials": 1,
      "running_trials": 0,
      "wrapper_exit_code": 130,
      "note": "Wrapper interrupted after Harbor progress showed 30m53s elapsed; no completed scorer result should be treated as final."
    }
  ]
}