Benchmarks
347 sources · 1,258 models × 3,038 benchmarks · 158,912 scoresAfterQuery's CrossSearch gives an agent 100 expert-written research requests that can only be answered by combining a task's local files (435 in all) with one to five live web sources the request doesn't name. Agents work in AfterQuery's Terminus 2 harness on Harbor, with one shell that has live internet but no search tool, and attempt each task three times. A GPT-6.1 Sol judge, blind to model identity, checks each answer against 2 to 20 binary rubric items (663 in all), and an attempt earns full credit only when every item passes. The score is pass^3, the share of tasks whose three attempts all earn full credit.
73rd strongest of 798 fitted benchmarks
Signal-to-noise +14.98
163rd hardest of 798 fitted benchmarks
Capability needed 166.9
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Gemini 4 Argon | high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 65,536-token response limit | 64.0 | |
| 2 | GPT-6 Astra | x-high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 128,000-token response limit | 60.0 | |
| 3 | GPT-6.1 Sol | x-high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 128,000-token response limit | 59.0 | |
| 4 | Claude Opus 5.5 | x-high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 128,000-token response limit | 57.0 | |
| 5 | Claude Fable 5.2 | high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 32,768-token response limit | 56.0 | |
| 6 | Muse Wonder | high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 32,768-token response limit | 42.0 | |
| 7 | GPT-5.6 Sol | x-high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 128,000-token response limit | 38.0 | |
| 8 | Grok 4.6 | x-high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 450,000-token response limit | 35.0 | |
| 9 | Claude Opus 5 | x-high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 128,000-token response limit | 34.0 | |
| 10 | Gemini 3.8 Flash | high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 65,536-token response limit | 29.0 | |
| 11 | Inkling 2 Small | thinking · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 262,144-token response limit; Tinker's 256K-context listing, asked for effort "high", whose numeric setting is undocumented | 23.0 | |
| 12 | Kimi K3 | high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 943,718-token response limit | 15.0 | |
| 13 | Sonnet 5 | x-high · AfterQuery CrossSearch harness: Terminus 2 on Harbor, one shell with live internet, no search tool and a view_image command; three attempts per task; 128,000-token response limit | 7.0 |
Benchmarks outside its family whose model rankings track it most reliably (Spearman ρ over at least 10 shared models).