Terminal-Bench 4
The TB3 contract, under the TB4 release name.
The platform resolves the TB4 release name to its TB3 audit profile. The task format, agent egress and verifier requirements are shared.
The conditions a task is audited under.
Read from src/infinity/profile.py, which resolves the profile once at load time so every check reads it from the bundle rather than each deciding for itself.
- Agent egress
- Open, exactly as TB3 -- the 4.0.0 task format carries the same egress contract.
- Verifier
- Separate container, no network. Treated as required, as under TB3.
- Difficulty evidence
- Not required, as under TB3. TB4 does not turn the difficulty check on.
TB4 uses the same audited requirements as TB3. Both release names resolve to one maintained contract.
Shared assurance requirements.
The verifier runs apart from the agent
Harbor makes a separate verifier sandbox opt-in. The audit treats it as required, because a verifier sharing a filesystem with the thing it is grading has no boundary left to test.
S11 in src/infinity/checks/separate.pyComplete required coverage before issuance
The report separates completed checks from skipped and errored stages. Every required check must complete before a mark can be issued.
grade() in platform/web/lib/certification.tsThe result names the revision it was graded under
Each result retains the methodology revision used for its audit, so later changes remain distinct from the original evidence.
METHODOLOGY_VERSION = 1.0 in platform/web/lib/brand.ts
Inspect the implementation sources.
Inspect the source definitions for the profile, checks and reference inventory.
src/infinity/profile.pyThe "tb4" and "terminal-bench-4" aliases, and the comment explaining themcorpus/tb5/README.md"The shipped release calls itself Terminal-Bench 4.0.0"
Submit the task. Get its recorded check evidence.
Tbench AI audits the task, environment, reference solution and verifier. Every completed audit retains its evidence report.