Dual-Arm Sim - Expert Honesty: leaderboard
micro1 Dual-Arm Sim’s October 30 release evaluates ten general-purpose models on four tasks, with 20 seeded layouts per task and 80 trials per model. Isaac Sim and Isaac Lab provide two Franka Panda arms. Each model receives images and sensors and returns structured actions, with 120 decisions per trial; simulation pauses during reasoning. Stage weights are 0, 20, 40, 60, 80 and 100, using ground-truth counts. Reasoning effort, temperature and exact API snapshots are unreported. The publisher explicitly hides its two embedded VLA rows as coming soon, so they are not admitted. Human reviewers rate trials and traces on five dimensions from 1 to 5. Overall weights execution at 40% and each other dimension at 15%. Read the published rounded overall rather than recomputing it. Question access: Published task descriptions, rubrics and sample episode videos; the exact 800-trial environments and full seeded task set are not released.
Metric: Published expert rating (1–5). Source: www.micro1.ai. Saturation forecast: Rough model projection: around 2026. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 5 | |
| 2 | Claude Sonnet 5.5 (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 5 | |
| 3 | Claude Fable 5.1 (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 4 | |
| 4 | Gemini 4 Argon (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 3.5 | |
| 5 | GPT-6 Astra (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 3.5 | |
| 6 | GPT-6.1 Sol (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 2.5 | |
| 7 | Gemini 3.8 Flash (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 2 | |
| 8 | Muse Wonder (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 1.5 | |
| 9 | Gemini 3.1 Pro (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 1 | |
| 10 | Grok 4.7 (Benchmark Sweep Run: source=Micro1DualArm; API; 120 decisions; effort and snapshot unreported) | 1 |
Interactive version: theaggregate.ai/benchmark?slug=dual-arm-sim-expert-honesty · How It Works · Data refreshed daily, snapshot 2026-11-06.