GPT-6 Astra

OpenAI#1 of 2 by best run
Compare models
Best success79%in Cline
Fastest run125sper task in Hermes Agent
Best harness
Cline
Graded runs
7 of 8 harnesses
Latest run
2026-10-15

Success rate

% of tasks passed by harness · higher is better
Cline
79.3%
Codex
75.9%
Hermes Agent
75.9%
Command Code
75.9%
Claude Code
72.4%
OpenCode
69.0%
Pi Agent
69.0%

Time per task

Seconds per task by harness · lower is better
Hermes Agent
125s
OpenCode
127s
Pi Agent
128s
Codex
149s
Command Code
159s
Cline
193s
Claude Code
198s

Runs

  1. 01Cline
    79%
    Time / task193s
  2. 02Hermes Agent
    76%
    Time / task125s
  3. 03Codex
    76%
    Time / task149s
  4. 04Command Code
    76%
    Time / task159s
  5. 05Claude Code
    72%
    Time / task198s
  6. 06OpenCode
    69%
    Time / task127s
  7. 07Pi Agent
    69%
    Time / task128s

1 harness not yet run. Pages fill in as new runs land