← GPT-5.6 Sol · API · Max effort
Evidenceacceptedindependent_lab
Terminal-Bench
Artificial Analysis — Terminal-Bench v2.1 (GPT-5.6 Sol max)
ev_2026_07_gpt56sol_terminal
Result
88%
Confidence A
- Tested
- 2026-07-09
- Runs
- 3
Configuration
How it was measured
- Exact model
- gpt-5.6-sol
- Reasoning
- max
- Agent harness
- Terminus-2 / AA Terminal-Bench v2.1
- Tool access
- terminal agent tools in e2b sandbox
- Prompting
- AA standard Terminal-Bench v2.1 protocol
STR stretch + skill deep-dive (long_autonomous_runs / terminal). Independent AA Terminal-Bench v2.1 under Terminus-2 harness.
Caveats
Known limitations
- Max effort row used for Build parity with INT/STR max config; xhigh scores slightly higher on AA TB 2.1.
- Do not mix with vendor Ultra multi-agent Terminal-Bench numbers.
Weighting
Evidence weight factors
The engine weights each evidence row by these factors (0–1) when fusing scores — higher is more trusted.
relevance0.95
quality1.00
recency1.00
health0.90
comparability0.85
reliability0.90
Benchmark
Terminal-Bench
Agentic terminal / tool-use suite measuring multi-step shell and environment task completion (Terminus-2 harness commonly cited).
Benchmark source →