Skip to main content
BenchLM
Data

Blueprint-Bench 2

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

Data verified

An agentic spatial reasoning benchmark reported as a normalized score.

Normalized connectivity score on Blueprint-Bench 2 — November 12, 2026

We mirror the published normalized connectivity score view for Blueprint-Bench 2. Gemini 4 Carbon leads the public snapshot at 58.6%, followed by Claude Fable 5.2 (56.6%) and Gemini 4 Argon (54.4%). We do not use these results to rank models overall.

1Closed

Gemini 4 Carbon

Google

58.6%
Overall 89.15Context 1M
2Closed

Claude Fable 5.2

Anthropic

56.6%
Overall 93.65Context 1M
3Closed

Gemini 4 Argon

Google

54.4%
Overall 79.03Context 1M

38 modelsMultimodal & GroundedCurrentDisplay onlyUpdated November 12, 2026

Normalized connectivity score table (38 models)

Normalized connectivity score results for Blueprint-Bench 2
RankModel / configurationScoreParameters (B)Open / closed
1Gemini 4 CarbonGoogle
58.6%
Not reportedClosed
2Claude Fable 5.2Anthropic
56.6%
Not reportedClosed
3Gemini 4 ArgonGoogle
54.4%
Not reportedClosed
4Claude Opus 5.5Anthropic
51.2%
Not reportedClosed
5GPT-6 AstraOpenAI
49.7%
Not reportedClosed
6GPT-6.1 SolOpenAI
43.2%
Not reportedClosed
7GPT-6.1 TerraOpenAI
42.1%
Not reportedClosed
8Claude Fable 5.1Anthropic
41.9%
Not reportedClosed
9Claude Sonnet 5.5Anthropic
39.3%
Not reportedClosed
10Claude Fable 5Anthropic
38.6%
Not reportedClosed
11Gemini 3.8 FlashGoogle
38.6%
Not reportedClosed
12GPT-6 SolOpenAI
36.9%
Not reportedClosed
13GPT-5.5OpenAI
36.2%
Not reportedClosed
14Gemini 3.5 FlashGoogle
33.6%
Not reportedClosed
15GPT-5.6 SolOpenAI
33.6%
Not reportedClosed
16Grok 4.6xAI
33.2%
Not reportedClosed
17Claude Haiku 5.5Anthropic
33.0%
Not reportedClosed
18Grok 4.7xAI
32.5%
Not reportedClosed
19Muse WonderMeta
31.7%
Not reportedClosed
20Gemini 3.6 FlashGoogle
31.2%
Not reportedClosed
21GPT-6 LunaOpenAI
31.2%
Not reportedClosed
22GPT-5.6 TerraOpenAI
30.8%
Not reportedClosed
23Claude Opus 5Anthropic
30.4%
Not reportedClosed
24Kimi K3Moonshot AI
29.5%
Not reportedOpen
25Grok 4.5xAI
27.3%
Not reportedClosed
26GPT-5.4OpenAI
27.1%
Not reportedClosed
27Gemini 3.1 ProGoogle
26.5%
Not reportedClosed
28Claude Sonnet 5Anthropic
24.9%
Not reportedClosed
29Claude Opus 4.7Anthropic
24.5%
Not reportedClosed
30GPT-5.6 LunaOpenAI
22.6%
Not reportedClosed
31Claude Opus 4.8Anthropic
14.5%
Not reportedClosed
32Claude Sonnet 4.6Anthropic
6.7%
Not reportedClosed
33Kimi K2.6Moonshot AI
3.9%
Not reportedOpen
34Gemini 3 FlashGoogle
0.0%
Not reportedClosed
35Grok 4.3xAI
0.0%
Not reportedClosed
36Gemini Robotics-ER 1.6Google
0.0%
Not reportedUnknown
37Claude Haiku 4.5Anthropic
0.0%
Not reportedClosed
38Grok 4.20xAI
0.0%
Not reportedClosed

How we show Blueprint-Bench 2

We mirror Andon Labs' Blueprint-Bench 2 model table as a display-only spatial-reasoning benchmark. Each agent processes 50 apartments sequentially, examines roughly 20 interior photographs per apartment, and draws a structured floor plan.

The composite measures room connectivity, graph degree and density, room and door counts, and orientation. Scores are normalized so the random baseline is 0 and a perfect result is 100. The human result covers 12 apartments, so it stays outside the model ranking.

Snapshot

38 model rows50 apartmentsHuman subset: 58.6%Display only
Blueprint-Bench 2 leaderboard and methodOriginal Blueprint-Bench paper

The published Blueprint-Bench 2 snapshot places Gemini 4 Carbon first at 58.6%. The third row is 4.2 points behind. The broader top-10 range is 20.0 points, so the table still separates the published systems.

38 models have been evaluated on Blueprint-Bench 2. The benchmark falls in the Multimodal & Grounded category. Blueprint-Bench 2 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Blueprint-Bench 2

Year

2026

Tasks

Spatial reasoning from blueprints

Format

Normalized score

Difficulty

Agentic spatial reasoning

Google reported Blueprint-Bench 2 in the Gemini 3.5 Flash launch comparison table. BenchLM stores it as a display-only multimodal and spatial-reasoning benchmark until Google publishes the full methodology page.

Gemini 3.5 Flash launch screenshotsPublic benchmark source

Freshness and provenance

Version

Blueprint-Bench 2 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 Blueprint-Bench 2 measure?

An agentic spatial reasoning benchmark reported as a normalized score.

Which model leads the published Blueprint-Bench 2 snapshot?

Gemini 4 Carbon currently leads the published Blueprint-Bench 2 snapshot with 58.6% normalized connectivity score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on Blueprint-Bench 2?

The November 12, 2026 snapshot contains 38 AI models.

Last updated: November 12, 2026 · 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 7,000+ subscribers.

One email each week. Unsubscribe anytime.