BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases7

Blueprint-Bench 2 leaderboard

As of 6 Nov 2026, the top score on Blueprint-Bench 2 is 58.6% by Gemini 4 Carbon (Google), ahead of Gemini 4 Argon at 54.5%, across 35 model configurations with a published result (Epoch AI Benchmarking Hub, result dated 28 Oct 2026). From photographs of a home the model must draw the floor plan, compared with the real one. Blueprint-Bench 2 is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 6 Nov 2026)

  • Gemini 4 Carbon holds the top Blueprint-Bench 2 score at 58.6% (Epoch AI Benchmarking Hub, 28 Oct 2026).
  • Gemini 4 Argon is second on Blueprint-Bench 2 at 54.5% (Epoch AI Benchmarking Hub, 30 Sep 2026).
  • Blueprint-Bench 2: Turning photos of rooms into floor plans (Epoch AI Benchmarking Hub).
  • 35 model configurations have a published Blueprint-Bench 2 result on BenchLeader (BenchLeader, 6 Nov 2026).
  • Blueprint-Bench 2 does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Turning photos of rooms into floor plans.

Published by
Epoch AI Benchmarking Hub
Category
Multimodal
Index weight
Reference only
Models
35
Data as of
6 Nov 2026
Newest result
28 Oct 2026

CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.

What the test looks like

From photographs of a home the model must draw the floor plan, compared with the real one.

How it is scored

Score, as published.

What to keep in mind

Vision models only; small set.

Top 15
  1. 1Gemini 4 Carbon58.6%
  2. 2Gemini 4 Argon54.5%
  3. 3Claude Opus 5.551.2%
  4. 4GPT-6 Astra49.7%
  5. 5GPT-6.1 Terra42.1%
  6. 6Claude Fable 5.141.9%
  7. 7Claude Sonnet 5.539.3%
  8. 8Claude Fable 538.6%
  9. 9Gemini 3.8 Flash38.6%
  10. 10GPT-6 Sol36.9%
  11. 11GPT-5.536.2%
  12. 12GPT-5.6 Sol33.6%
  13. 13Gemini 3.5 Flash33.6%
  14. 14Grok 4.633.2%
  15. 15Claude Haiku 5.533.0%
35 of 35
#
1Gemini 4 Carbon58.6%–2026-10-28
2Gemini 4 Argon54.5%63.42026-09-30
3Claude Opus 5.551.2%63.42026-09-22
4GPT-6 Astra49.7%68.22026-09-03
5GPT-6.1 Terra42.1%–2026-10-26
6Claude Fable 5.141.9%66.42026-09-01
7Claude Sonnet 5.539.3%–2026-09-28
8Claude Fable 538.6%66.52026-06-09
9Gemini 3.8 Flash38.6%54.42026-09-02
10GPT-6 Sol36.9%–2026-09-22
11GPT-5.536.2%61.72026-04-23
12GPT-5.6 Sol33.6%55.02026-07-09
13Gemini 3.5 Flash33.6%48.62026-05-19
14Grok 4.633.2%57.12026-08-12
15Claude Haiku 5.533.0%–2026-10-07
16Grok 4.732.5%–2026-09-21
17Gemini 3.6 Flash31.2%49.52026-07-21
18GPT-6 Luna31.2%–2026-09-22
19GPT-5.6 Terra30.8%51.52026-07-09
20Claude Opus 530.4%65.22026-07-24
21Kimi K329.5%59.12026-07-16
22Grok 4.527.3%59.82026-07-08
23GPT-5.427.1%58.02026-03-05
24Gemini 3.1 Pro26.5%60.52026-02-19
25Claude Sonnet 524.9%55.92026-06-30
26Claude Opus 4.724.5%63.12026-04-16
27GPT-5.6 Luna22.6%47.62026-07-09
28Claude Opus 4.814.5%60.02026-05-28
29Claude Sonnet 4.66.7%58.22026-02-17
30Kimi K2.63.9%55.22026-04-20
31Gemini 3 Flash0.0%56.92025-12-17
32Grok 4.30.0%54.52026-04-17
33Claude Haiku 4.50.0%47.62025-10-15
34Grok 4.200.0%–2026-02-17
35Gemini Robotics-ER 1.60.0%–2026-04-14

More multimodal benchmarks · All LLM benchmarks

Related benchmarks

Top head-to-heads on Blueprint-Bench 2

Cite as: BenchLeader, “Blueprint-Bench 2 leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/blueprint_bench_2.

Blueprint-Bench 2: questions

What does Blueprint-Bench 2 measure?
From photographs of a home the model must draw the floor plan, compared with the real one. Scores are reported in percent of tasks solved; higher is better.
What is the top score on Blueprint-Bench 2?
As of 6 Nov 2026, the highest score on Blueprint-Bench 2 is 58.6% by Gemini 4 Carbon (result dated 28 Oct 2026), ahead of Gemini 4 Argon at 54.5%. Scores come from Epoch AI Benchmarking Hub and are re-read daily.
Which AI model leads Blueprint-Bench 2?
Gemini 4 Carbon leads Blueprint-Bench 2 with 58.6% as of 6 Nov 2026 (result dated 28 Oct 2026), ahead of Gemini 4 Argon at 54.5%.
How many models have Blueprint-Bench 2 results?
35 model configurations have a Blueprint-Bench 2 result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
Who runs Blueprint-Bench 2 and how often is it updated?
Blueprint-Bench 2 is published by Epoch AI Benchmarking Hub. BenchLeader re-reads the published results once a day and records the date each result was published.
Does Blueprint-Bench 2 count toward the BenchLeader Index?
No. Blueprint-Bench 2 is shown for reference but left out of the composite index.