Benchmark Almanac

BabyAI

GPT-4 (snapshot unspecified) 56.2% tops its leaderboard Claude 2 37.5% leads the models we track
Not in model cards

Agents follow compositional English instructions to navigate rooms, open doors, and move objects in a grid world. The environment checks whether each instruction was completed, and evaluation reports task success rates.

BabyAI environments now live in the Minigrid library. The original repository retains the published baselines and evaluation scripts.

Made by
Mila
Released
October 2018
Kind
Simulated robots
Level
Everyday
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Images → Actions
Interaction
Agentic
Environment
MiniGrid
Items from
Generated
Refresh
One fixed set
Run by
Anyone
In views
Agents, Vision, Robotics
Strongest model tested
Claude 2 #192 on the capability index
Last model added
December 2023
Scores refreshed
November 20th, 2026

In the authors’ words

Allowing humans to interactively train artificial agents to understand language instructions is desirable for both practical and scientific reasons, but given the poor data efficiency of the current learning methods, this goal may require substantial research efforts. Here, we introduce the BabyAI research platform to support investigations towards including humans in the loop for grounded language learning. The BabyAI platform comprises an extensible suite of 19 levels of increasing difficulty. The levels gradually lead the agent towards acquiring a combinatorially rich synthetic language which is a proper subset of English. The platform also provides a heuristic expert agent for the purpose of simulating a human teacher. We report baseline results and estimate the amount of human involvement that would be required to train a neural network-based agent on some of the BabyAI levels. We put forward strong evidence that current deep learning methods are not yet sufficiently sample efficient when it comes to learning a language with compositional properties.

Abstract of BabyAI: A Platform to Study the Sample Efficiency of Grounded Language Learning (arXiv:1810.08272, CC0)
13
Entries on AgentBoard’s leaderboard, 5 models we track

GPT-4 (snapshot unspecified) tops the leaderboard with 56.2% Claude 2 leads with 37.5% among models released by December 2023

Best score per entry from AgentBoard’s leaderboard, read on November 20th, 2026. AgentBoard reports the success rate across its 112 BabyAI episodes. 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. The best model we track is Claude 2, at 37.5%. This view shows only models we track, each at its best.

The newest model with a score here came out in December 2023, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-4 (snapshot unspecified) 56.2%
  2. 2 GPT-3.5-Turbo (snapshot unspecified) 39.3%
  3. 3 Claude 2 37.5%
  4. 4 GPT-3.5-Turbo-16k 33.9%
  5. 5 DeepSeek LLM 67B 22.3%
  6. 6 Llama-2-70b-chat-hf 19.6%
  7. 7 CodeLlama-13b 17%
  8. 8 davinci-003 14.3%
  9. 9 Mistral-7B-Instruct-v0.2 14.3%
  10. 10 Code Llama 34B Instruct 13.4%
  11. 11 Lemur-70b 9.80%
  12. 12 Llama-2-13b-chat-hf 6.20%
  13. 13 Vicuna-13b-16k 5.40%

Source: AgentBoard’s leaderboard

#EntryScoreSettingDateSource
1 GPT-4 (snapshot unspecified) 56.2% AgentBoard Official leaderboard
2 GPT-3.5-Turbo (snapshot unspecified) 39.3% AgentBoard Official leaderboard
3 Claude 2 37.5% AgentBoard Jul 2023 Official leaderboard
4 GPT-3.5-Turbo-16k 33.9% AgentBoard Official leaderboard
5 DeepSeek LLM 67B 22.3% AgentBoard Nov 2023 Official leaderboard
6 Llama-2-70b-chat-hf 19.6% AgentBoard Official leaderboard
7 CodeLlama-13b 17% AgentBoard Official leaderboard
8 davinci-003 14.3% AgentBoard Nov 2022 Official leaderboard
9 Mistral-7B-Instruct-v0.2 14.3% AgentBoard Dec 2023 Official leaderboard
10 Code Llama 34B Instruct 13.4% AgentBoard Aug 2023 Official leaderboard
11 Lemur-70b 9.80% AgentBoard Official leaderboard
12 Llama-2-13b-chat-hf 6.20% AgentBoard Official leaderboard
13 Vicuna-13b-16k 5.40% AgentBoard Official leaderboard
#ModelScoreSettingDateSource
1 Claude 2 37.5% AgentBoard Jul 2023 Official leaderboard
2 DeepSeek LLM 67B 22.3% AgentBoard Nov 2023 Official leaderboard
3 davinci-003 14.3% AgentBoard Nov 2022 Official leaderboard
4 Mistral-7B-Instruct-v0.2 14.3% AgentBoard Dec 2023 Official leaderboard
5 Code Llama 34B Instruct 13.4% AgentBoard Aug 2023 Official leaderboard

2 models have held number one on AgentBoard’s leaderboard since 2023

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

OpenAI Anthropic
AgentBoard leaderboard Not tested since Dec 2023
Claude 2 37.5%

Source: AgentBoard’s 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 AgentBoard’s leaderboard Best so far
0% 20% 40% 60% 80% 100% 2023 2024 Near the ceiling (85%) Claude 2 davinci-003 0% 20% 40% 60% 80% 100% 2023 2024 Near the ceiling (85%) Claude 2 davinci-003

Source: AgentBoard’s 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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in December 2023, for Mistral-7B-Instruct-v0.2. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (13 entries).
  • Pass Models scored 5 models we track, plus 8 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in December 2023, for Mistral-7B-Instruct-v0.2.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude 2 (#192).
  • Pass Headroom The best is 37.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on AgentBoard: 5 rows for models we track and 8 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