Skip to main content
BenchLM

Scale Labs FORTRESS (FORTRESS)

We show this table for reference; we do not rank on it.

Data verified 36 confirmed releases in the last 30 daysFollow model changes

A Scale Labs public leaderboard mirrored as display-only reference data. It does not affect BenchLM rankings. The catalog still links this board, but its public route was unavailable during the latest refresh; these are the last successfully captured rows.

Scale score on FORTRESS — 2026-09-28 last available snapshot

We mirror the published scale score view for FORTRESS. DeepSeek-R1 leads the public snapshot at 74.4%, followed by glm-4p5-air (63.2%) and Gemini 2.5 Pro (06-05) (61.7%). We do not use these results to rank models overall.

68 modelsKnowledgeCurrentDisplay onlyUpdated 2026-09-28 last available snapshot

Scale score table (68 models)

Score
1
DeepSeek-R1DeepSeek · Open weight
74.4%
2
63.2%
4
61.4%
5
DeepSeek V3.1DeepSeek · Open weight
60.5%
6
59.6%
7
GPT-4.1 miniOpenAI · Closed
59.2%
8
56.4%
9
56.1%
10
kimi-k2-instructMoonshot AI
55.5%
12
Gemini 1.5 ProGoogle · Closed
53.9%
13
GPT-4.1OpenAI · Closed
53.0%
15
50.6%
16
GPT-4o miniOpenAI · Closed
48.1%
17
GPT-4oOpenAI · Closed
47.2%
18
44.8%
19
44.2%
21
Kimi K2.5Moonshot AI · Open weight
41.1%
22
Llama 4 MaverickMeta · Open weight
40.1%
23
38.0%
24
30.4%
25
30.4%
26
o3-miniOpenAI · Closed
30.1%
28
28.2%
29
Claude Opus 4Anthropic
27.6%
30
Kimi K3Moonshot AI · Closed
27.3%
31
kimi-k3 maxMoonshot AI
26.6%
33
25.7%
34
inkling xhighThinking Machines
25.0%
35
24.8%
36
24.4%
38
o4-miniOpenAI
21.5%
39
Llama 3.1 405BMeta · Open weight
20.6%
41
Muse SparkMeta · Closed
20.2%
42
GPT-5.6 SolOpenAI · Closed
20.1%
43
o1OpenAI · Closed
19.4%
44
18.2%
46
GPT-OSS 20BOpenAI · Open weight
17.6%
47
GPT-5.2OpenAI · Closed
17.5%
48
17.0%
49
GPT-5 miniOpenAI · Closed
17%
50
Claude Sonnet 4.5Anthropic · Closed
17.0%
51
Claude Opus 5Anthropic · Closed
16.8%
52
16.3%
53
16.1%
54
o3OpenAI · Closed
16.0%
56
GPT-5.4 ProOpenAI · Closed
14.8%
58
Fable-5.1Anthropic
13.7%
59
Claude Opus 4.5Anthropic · Closed
13.6%
60
GPT-6 SolOpenAI · Closed
13.5%
61
Claude Opus 5.5Anthropic · Closed
13.2%
63
Claude 3.5 SonnetAnthropic · Closed
13.0%
64
Claude Sonnet 4.5 ThinkingAnthropic · Closed
12.8%
65
Muse Spark 1.1Meta · Closed
12.4%
66
GPT-6 LunaOpenAI · Closed
10.8%
67
Claude Opus 4.5 ThinkingAnthropic · Closed
9.6%
68
GPT-OSS 120BOpenAI · Open weight
8.2%

How to read this leaderboard

Compare the published configurations as complete evaluation systems. The source can combine a base model, agent scaffold, tools, budget, and inference setting in each result.

Operator receipt: 68 sourced rows are currently displayable on this page; the leading published row is DeepSeek-R1 at 74.4%.

Honest limit: This Scale table is display-only context, not benchmark provenance or a weighted model-only comparison. The catalog still links this board, but its public route was unavailable during the latest refresh; these are the last successfully captured rows.

How BenchLM shows FORTRESS

BenchLM mirrors 68 published rows from Scale Labs’ public FORTRESS leaderboard, captured on 2026-09-28 last available snapshot. The catalog still links this board, but its public route was unavailable during the latest refresh; these are the last successfully captured rows.

The table is display only. It is useful context for a published agent or model configuration, but it does not enter BenchLM’s overall or category rankings.

Snapshot

68 published rowsScale Labs sourceSource currently unavailableDisplay only

The published FORTRESS snapshot places DeepSeek-R1 first at 74.4%. The third row is 12.7 points behind. The broader top-10 range is 18.9 points, so the table still separates the published systems.

68 models have been evaluated on FORTRESS. The benchmark falls in the Knowledge category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. FORTRESS is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About FORTRESS

Year

2026

Tasks

68 published rows

Format

Published Scale leaderboard score

Difficulty

External agent and model evaluation

BenchLM mirrors 68 published rows from the FORTRESS public table captured on 2026-09-28 last available snapshot.

Freshness and provenance

Version

FORTRESS 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does FORTRESS measure?

A Scale Labs public leaderboard mirrored as display-only reference data. It does not affect BenchLM rankings. The catalog still links this board, but its public route was unavailable during the latest refresh; these are the last successfully captured rows.

Which model leads the published FORTRESS snapshot?

DeepSeek-R1 currently leads the published FORTRESS snapshot with 74.4% scale score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on FORTRESS?

The 2026-09-28 last available snapshot snapshot contains 68 AI models.

Last updated: 2026-09-28 last available snapshot · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.