Claim · agents

GLM 5.1on Terminal-Bench 2

59.3 to 59.3 the range across every scoreable measurement we hold

0.0 points between the highest and the lowest

59.3 the number the board uses: the lower median, a real row with its own provenance

1 declared configuration, from 1 runner

0 / 4 rows whose cited page still shows this figure, refetched and checked by us

All 1 scoreable measurements here come from one runner under one declared configuration, so nothing on this page explains why they differ. That is the finding.

3 of 4 cited pages no longer show the figure we publish. The page was refetched and split at its own structural boundaries, and no block that names this model carries this number. That does not make the measurement wrong: a leaderboard is a living page and a figure recorded when it was there can be gone by the time you follow the link. It does mean the citation, as printed, no longer supports the number, and this is the only place we know of that says so about its own sources.

What the vendor said, and what someone else measured

Matched on which variant of the benchmark was sat, and on nothing else: the vendor ran it on their harness and the independent party on theirs, and that difference is the comparison rather than an obstacle to it. A vendor number with no independent run of the same variant carries no delta, because a number that cannot be compared honestly must not be compared.

Variant	Vendor claims	Independently measured	Difference
2.1	62.0	59.3 from 1 run	+2.7

Vendor numbers nobody has independently checked

Published by the model's own maker, on a variant of this benchmark that no independent party we hold has run. They are recorded and they are shown. They can never move a rank.

Variant	Claimed
2.0 best self-reported Claude Code	69.0
2.0 Terminus-2	63.5

Every measurement

4 rows. Best first. the board takes the LOWER MEDIAN of the scoreable measurements, which is a real row with its own interval and provenance rather than an interpolated average. a vendor self-report is shown beside the independent numbers and is never one of them; it cannot move a rank.

Score	Provenance	Configuration	Who ran it	Where to check it
69.0 / 100 as published: 69.0%	CLAIMED shown, never scored	glm-5.1 harness Claude Code effort not disclosed variant 2.0 best self-reported Claude Code	Z.AI via Model benchmark dossier facts, not expression	checked: this figure is not on the page today2026-08-30GLM-5.1 huggingface.co
63.5 / 100 as published: 63.5%	CLAIMED shown, never scored	glm-5.1 harness model-card evaluation effort not disclosed variant 2.0 Terminus-2	Z.AI via Model benchmark dossier facts, not expression	checked: this figure is not on the page today2026-08-30GLM-5.1 huggingface.co
62.0 / 100 as published: 62.0%	CLAIMED shown, never scored	glm-5.1 harness later GLM-5.2 comparison effort not disclosed variant 2.1	Z.AI via Model benchmark dossier facts, not expression	checked: found, but not in a block naming only this model2026-08-30On standard coding benchmarks, GLM-5.2 is the strongest open-source model, improving on GLM-5.1 by a wide margin: 81.0 vs. 62.0 on Terminal-Bench 2.1 and 62.1 v github.com
59.3 / 100 as published: 59.3%	INDEPENDENT counts toward the ranking	glm-5.1 harness NeMo Evaluator SDK; Harbor effort model-card defaults variant 2.1	NVIDIA via Model benchmark dossier facts, not expression	not checked: the page is a format we cannot read2026-08-30 research.nvidia.com

← The Forge Rankings

Snapshot 2026-08-30-62e4e85f043b, manifest dc46131315046f1e, as of 2026-08-30. where only one party has measured a model on a benchmark, this page says so rather than presenting a single number as a reconciliation.