Benchmark Almanac

PhAIL DROID: Everyday tasks

Cosmos3-Nano 68.1% leads since October 2026
Not in model cards

Policies run 286 language-instructed tabletop tasks worded as DROID annotations on a real Franka FR3. Operators mark stages reached within 240 seconds.

Reliability comes from an item-response ranking model fitted over all tasks, with bootstrap 95% ranges. Absent from a June 2026 snapshot.

Released
October 2026
Kind
Real robots
Level
Everyday
Scored
Graded by people
Answers
Agent task
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
7
Entries on its own leaderboard, 6 models we track

Cosmos3-Nano leads with 68.1% among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. 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 newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 68.1 Cosmos3-Nano
  2. 62.5 π0.5
  3. 62.3 GR00T N1.7
  4. 61.7 FLUX 3 Action
  5. 55.9 GPT-6 Astra
  6. 44.5 MolmoAct 2
  7. 38.9 G0.5
  1. 1 Cosmos3-Nano 68.1%
  2. 2 π0.5 62.5%
  3. 3 GR00T N1.7 62.3%
  4. 4 FLUX 3 Action 61.7%
  5. 5 GPT-6 Astra 55.9%
  6. 6 MolmoAct 2 44.5%
  7. 7 G0.5 38.9%

Source: Official PhAIL DROID: Everyday tasks leaderboard

#EntryScoreSettingDateSource
1 Cosmos3-Nano 68.1% May 2026 Official leaderboard
2 π0.5 62.5% Apr 2025 Official leaderboard
3 GR00T N1.7 62.3% Apr 2026 Official leaderboard
4 FLUX 3 Action 61.7% Sep 2026 Official leaderboard
5 GPT-6 Astra 55.9% Sep 2026 Official leaderboard
6 MolmoAct 2 44.5% May 2026 Official leaderboard
7 G0.5 38.9% Official leaderboard

Reliability vs. speed at 240 seconds

The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

40% 50% 60% 70% 80% 00.511.52 Speed (tasks/min) Reliability MolmoAct 2: Reliability 44.5%, speed 1.3 tasks/min GPT-6 Astra: Reliability 55.9%, speed 0.41 tasks/min FLUX 3 Action: Reliability 61.7%, speed 1.44 tasks/min G0.5: Reliability 38.9%, speed 1 tasks/min. Not in our model list π0.5: Reliability 62.5%, speed 1.76 tasks/min GR00T N1.7: Reliability 62.3%, speed 1.79 tasks/min Cosmos3-Nano: Reliability 68.1%, speed 1.23 tasks/min Cosmos3-Nanoπ0.5GR00T N1.7 40% 50% 60% 70% 80% 00.511.52 Speed (tasks/min) Reliability MolmoAct 2: Reliability 44.5%, speed 1.3 tasks/min GPT-6 Astra: Reliability 55.9%, speed 0.41 tasks/min FLUX 3 Action: Reliability 61.7%, speed 1.44 tasks/min G0.5: Reliability 38.9%, speed 1 tasks/min. Not in our model list π0.5: Reliability 62.5%, speed 1.76 tasks/min GR00T N1.7: Reliability 62.3%, speed 1.79 tasks/min Cosmos3-Nano: Reliability 68.1%, speed 1.23 tasks/min Cosmos3-Nanoπ0.5GR00T N1.7
ModelSettingReliabilitySpeed
Cosmos3-Nano 68.1% 1.23 tasks/min
π0.5 62.5% 1.76 tasks/min
GR00T N1.7 62.3% 1.79 tasks/min
FLUX 3 Action 61.7% 1.44 tasks/min
GPT-6 Astra 55.9% 0.41 tasks/min
MolmoAct 2 44.5% 1.3 tasks/min
G0.5 38.9% 1 tasks/min

Number one over time

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

NVIDIA
Own leaderboard 1 leader
Cosmos3-Nano 68.1%

Source: Official PhAIL DROID: Everyday tasks 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% 2025 2026 Near the ceiling (85%) Cosmos3-Nano π0.5 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Cosmos3-Nano π0.5

Source: Official PhAIL DROID: Everyday tasks 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 103 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%