Benchmark Almanac

Embodied First Exam

Claude Fable 5.2 49.6% leads since November 2026
Not in model cards

Coding agents answer frozen robot tasks from many benchmark suites with a program. Each program is verified in a separate sandbox, run on Harbor.

Released October 19th, 2026. About 220 GitHub stars in its first weeks.

Released
October 2026
Kind
Simulated robots
Level
Everyday
Scored
Checked automatically
Answers
Code
Tasks
All public
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last model added
November 2026
Scores refreshed
November 12th, 2026
9
Models scored on its own leaderboard

Claude Fable 5.2 leads with 49.6%

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

  1. 49.6 Claude Fable 5.2
  2. 34.4 GPT-6 Astra
  3. 25.0 Claude Opus 5.5
  4. 24.1 Gemini 4 Argon
  5. 10.1 Kimi K3.1 Pro
  6. 6.36 Gemini 3.8 Flash
  7. 3.73 Qwen 3.8 Max
  8. 3.07 DeepSeek V4.1 Flash
  9. 2.41 GLM-5.3-Flash
  1. 1 Claude Fable 5.2 standard; high 49.6%
  2. 2 GPT-6 Astra standard; high 34.4%
  3. 3 Claude Opus 5.5 standard; high 25%
  4. 4 Gemini 4 Argon standard; high 24.1%
  5. 5 Kimi K3.1 Pro standard; high 10.1%
  6. 6 Gemini 3.8 Flash standard; high 6.36%
  7. 7 Qwen 3.8 Max standard; high 3.73%
  8. 8 DeepSeek V4.1 Flash standard; high 3.07%
  9. 9 GLM-5.3-Flash standard; high 2.41%

Source: Official Embodied First Exam leaderboard

#ModelScoreSettingDateSource
1 Claude Fable 5.2 49.6% standard; high, Claude Code Nov 2026 Official leaderboard
2 GPT-6 Astra 34.4% standard; high, Codex Sep 2026 Official leaderboard
3 Claude Opus 5.5 25% standard; high, Claude Code Sep 2026 Official leaderboard
4 Gemini 4 Argon 24.1% standard; high, Antigravity Sep 2026 Official leaderboard
5 Kimi K3.1 Pro 10.1% standard; high, Claude Code Oct 2026 Official leaderboard
6 Gemini 3.8 Flash 6.36% standard; high, Antigravity Sep 2026 Official leaderboard
7 Qwen 3.8 Max 3.73% standard; high, Claude Code Aug 2026 Official leaderboard
8 DeepSeek V4.1 Flash 3.07% standard; high, dsh Sep 2026 Official leaderboard
9 GLM-5.3-Flash 2.41% standard; high, Claude Code Aug 2026 Official leaderboard

The best score, 49.6%, costs $13 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Qwen DeepSeek Anthropic Google DeepMind Cost frontier One model at each effort level
0% 10% 20% 30% 40% 50% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5.2, standard; high Claude Opus 5.5, standard; high Qwen 3.8 Max, standard; high 0% 10% 20% 30% 40% 50% $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5.2, standard; high Claude Opus 5.5, standard; high

Source: Official Embodied First Exam leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 standard; high Claude Code 49.6% $13
GPT-6 Astra standard; high Codex 34.4% $25
Claude Opus 5.5 standard; high Claude Code 25% $5.59
Gemini 4 Argon standard; high Antigravity 24.1% $5.29
Kimi K3.1 Pro standard; high Claude Code 10.1% $8.00
Gemini 3.8 Flash standard; high Antigravity 6.36% $7.27
Qwen 3.8 Max standard; high Claude Code 3.73% $3.81
DeepSeek V4.1 Flash standard; high dsh 3.07% $0.42
GLM-5.3-Flash standard; high Claude Code 2.41% $0.62

2 models have held number one on the official leaderboard since its release in October 2026

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

OpenAI Anthropic
Own leaderboard 2 leaders
Claude Fable 5.2 49.6%

Source: Official Embodied First Exam leaderboard

Scores over time

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

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Claude Fable 5.2 GPT-6 Astra Gemini 3.8 Flash 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) Claude Fable 5.2 GPT-6 Astra

Source: Official Embodied First Exam leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Robotics

All 104 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%
  • 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 29%
  • RoboLab In use Zero-shot DROID Franka policies in Isaac Lab on 120 tabletop tasks at three instruction levels, 10 episodes each. Success rate. FLUX 3 Action 42.9%
  • 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%
  • Moment finding (Google DeepMind) In use Models find the moment in a robot video when an event happens. Scored by accuracy and mean absolute time distance in seconds. Gemini Robotics ER 2 91.3%