Benchmark Almanac

Embodied Agent Interface

o1-preview 70.3 leads since October 2024
Not in model cards

Models interpret goals, decompose them into subgoals, sequence actions, and predict state transitions in BEHAVIOR and VirtualHome. The evaluator checks task completion and distinguishes hallucinations, invalid actions, and planning errors.

The four modules have separate metrics. Running the BEHAVIOR track requires iGibson assets.

Released
October 2024
Kind
Embodied reasoning
Level
Everyday
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text, Actions
Interaction
Agentic
Environment
VirtualHome, BEHAVIOR
Refresh
One fixed set
Run by
Anyone
Based on
BEHAVIOR-1K
In views
Agents, Robotics
Strongest model tested
o1-mini #142 on the capability index
Last model added
September 2024
Scores refreshed
November 20th, 2026

In the authors’ words

We aim to evaluate Large Language Models (LLMs) for embodied decision making. While a significant body of work has been leveraging LLMs for decision making in embodied environments, we still lack a systematic understanding of their performance because they are usually applied in different domains, for different purposes, and built based on different inputs and outputs. Furthermore, existing evaluations tend to rely solely on a final success rate, making it difficult to pinpoint what ability is missing in LLMs and where the problem lies, which in turn blocks embodied agents from leveraging LLMs effectively and selectively. To address these limitations, we propose a generalized interface (Embodied Agent Interface) that supports the formalization of various types of tasks and input-output specifications of LLM-based modules. Specifically, it allows us to unify 1) a broad set of embodied decision-making tasks involving both state and temporally extended goals, 2) four commonly-used LLM-based modules for decision making: goal interpretation, subgoal decomposition, action sequencing, and transition modeling, and 3) a collection of fine-grained metrics which break down evaluation into various types of errors, such as hallucination errors, affordance errors, various types of planning errors, etc. Overall, our benchmark offers a comprehensive assessment of LLMs' performance for different subtasks, pinpointing the strengths and weaknesses in LLM-powered embodied AI systems, and providing insights for effective and selective use of LLMs in embodied decision making.

Abstract of Embodied Agent Interface: Benchmarking LLMs for Embodied Decision Making (arXiv:2410.07166, CC0)
18
Entries on its own leaderboard, 13 models we track

o1-preview leads with 70.3 among models released by September 2024 o1-preview leads with 70.3 among models released by September 2024

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The leaderboard gives a separate overall score for each environment, so the headline averages BEHAVIOR and VirtualHome equally. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 o1-preview 70.3
  2. 2 Claude 3.5 Sonnet 64.5
  3. 3 GPT-4o May 2024 60.3
  4. 4 Claude 3 Opus 60.1
  5. 5 o1-mini 57.7
  6. 6 Gemini 1.5 Pro (snapshot unspecified) 57.0
  7. 7 Gemini 1.5 Flash May 2024 54.0
  8. 8 GPT-4 Turbo Apr 2024 53.0
  9. 9 Claude 3 Sonnet 52.9
  10. 10 Mistral Large 52.5
  11. 11 Llama3 70B Instruct 47.3
  12. 12 Mixtral 8x22B MoE 45.0
  13. 13 Cohere Command R+ 44.3
  14. 14 Claude 3 Haiku 44.0
  15. 15 Gemini 1.0 Pro 37.2
  16. 16 Command R (03-2024) 33.0
  17. 17 GPT-3.5 Turbo Jan 2024 31.8
  18. 18 Llama3 8B Instruct 25.8

Source: Official Embodied Agent Interface leaderboard

#EntryScoreSettingDateSource
1 o1-preview 70.3 Sep 2024 Official leaderboard
2 Claude 3.5 Sonnet 64.5 Jun 2024 Official leaderboard
3 GPT-4o May 2024 60.3 May 2024 Official leaderboard
4 Claude 3 Opus 60.1 Feb 2024 Official leaderboard
5 o1-mini 57.7 Sep 2024 Official leaderboard
6 Gemini 1.5 Pro (snapshot unspecified) closed weights 57.0 Official leaderboard
7 Gemini 1.5 Flash May 2024 54.0 May 2024 Official leaderboard
8 GPT-4 Turbo Apr 2024 53.0 Apr 2024 Official leaderboard
9 Claude 3 Sonnet 52.9 Feb 2024 Official leaderboard
10 Mistral Large 52.5 Feb 2024 Official leaderboard
11 Llama3 70B Instruct open weights 47.3 Official leaderboard
12 Mixtral 8x22B MoE open weights 45.0 Official leaderboard
13 Cohere Command R+ open weights 44.3 Official leaderboard
14 Claude 3 Haiku 44.0 Mar 2024 Official leaderboard
15 Gemini 1.0 Pro 37.2 Dec 2023 Official leaderboard
16 Command R (03-2024) 33.0 Mar 2024 Official leaderboard
17 GPT-3.5 Turbo Jan 2024 31.8 Jan 2024 Official leaderboard
18 Llama3 8B Instruct open weights 25.8 Official leaderboard
#ModelScoreSettingDateSource
1 o1-preview 70.3 Sep 2024 Official leaderboard
2 Claude 3.5 Sonnet 64.5 Jun 2024 Official leaderboard
3 GPT-4o May 2024 60.3 May 2024 Official leaderboard
4 Claude 3 Opus 60.1 Feb 2024 Official leaderboard
5 o1-mini 57.7 Sep 2024 Official leaderboard
6 Gemini 1.5 Flash May 2024 54.0 May 2024 Official leaderboard
7 GPT-4 Turbo Apr 2024 53.0 Apr 2024 Official leaderboard
8 Claude 3 Sonnet 52.9 Feb 2024 Official leaderboard
9 Mistral Large 52.5 Feb 2024 Official leaderboard
10 Claude 3 Haiku 44.0 Mar 2024 Official leaderboard
11 Gemini 1.0 Pro 37.2 Dec 2023 Official leaderboard
12 Command R (03-2024) 33.0 Mar 2024 Official leaderboard
13 GPT-3.5 Turbo Jan 2024 31.8 Jan 2024 Official leaderboard

Rankings by part

We read 16 parts from the benchmark’s own leaderboard in 2 groups: behavior and virtualhome. Each ranks models on one part of the test. o1-preview leads 10 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Overall Performance

13 models we track have a score here.

  1. 1 o1-preview 74.9
  2. 2 Claude 3.5 Sonnet 64.2
  3. 3 Claude 3 Opus ↑1 60.4
  4. 4 GPT-4o May 2024 ↓1 59.8
  5. 5 o1-mini 57.5
  6. 6 Claude 3 Sonnet ↑2 55.1
  7. 7 Gemini 1.5 Flash May 2024 ↓1 52.1
  8. 8 Mistral Large ↑1 50.4
  9. 9 GPT-4 Turbo Apr 2024 ↓2 49.6
  10. 10 Claude 3 Haiku 41.6

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard Not tested since Oct 2024
o1-preview 70.3

Source: Official Embodied Agent Interface leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
30.0 40.0 50.0 60.0 70.0 80.0 2024 30.0 40.0 50.0 60.0 70.0 80.0 2024

Source: Official Embodied Agent Interface leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Robotics

All 172 in Robotics
  • ERQA In use 400 multiple-choice visual questions from robot and egocentric footage on spatial, trajectory, action and state reasoning. Accuracy. Gemini Robotics ER 2 78.5%
  • SimplerEnv In use Real-to-sim replicas of Google Robot and WidowX/Bridge setups. Policies scored by success rate on pick, move, drawer and Bridge tasks. Praxis-1 80%
  • RoboLab In use Zero-shot DROID Franka policies in Isaac Lab on 120 tabletop tasks at three instruction levels, 10 episodes each. Success rate. Praxis-1 46.2%
  • RoboDojo In use Generalist manipulation policies run 42 Isaac Sim tasks and 18 remote real-robot tasks. Scored by task score and success rate. GPT-6 Astra 27.6%
  • Progress classification (Google DeepMind) In use Models assign each frame of a robot video to one of five task-progress bands from 0 to 100 percent. Scored by accuracy. Gemini Robotics ER 2 57.4%
  • Human proximity (Google DeepMind robotics) In use Models judge from robot camera views whether a human is within 1 meter, a spatial awareness check for safe robot behavior. Scored by accuracy. Gemini Robotics ER 2 93%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in September 2024, for o1-mini. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (18 entries).
  • Pass Models scored 13 models we track, plus 5 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in September 2024, for o1-mini.
  • Note Frontier models None of the top 10 has a score. The strongest tested is o1-mini (#142).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on embodied-agent-interface.github.io: 13 rows for models we track and 5 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check