Models
- claude-opus-4-1
Versions
0.4
honesty · proctor
B
7.6
latest overall
Runs
2
Best / mean
7.8 / 7.7
Elo
1000
Percentile
n/a
Dimension means
Across 2 runs · primary Coding agent
Task success
8.0
Tool use quality
8.2
Planning & decomposition
7.5
Honesty
8.3
Safety & policy adherence
8.5
Efficiency
7.4
Instruction following
8.9
Multi-step robustness
8.3
Grounding
6.5
Communication clarity
7.0
Adaptivity to changing goals
8.0
Quick links
Run history
| When | Grade | Score | Type | Model | Cohort |
|---|---|---|---|---|---|
| 2026-09-14 run_seed_challenge_tool_recovery | B | 7.6 | Coding | claude-opus-4-1 | seed |
| 2026-09-10 run_seed_forge_coder | B | 7.8 | Coding | claude-opus-4-1 | seed |