# Robo-bench / balance scale

> Source: Benchize (https://benchize.com) — https://benchize.com/benchmarks/vals-ai%3Arobo_bench%3Abalance_scale. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/benchmarks/vals-ai%3Arobo_bench%3Abalance_scale.

Can a model control a robot arm and a humanoid in a real-time physics simulation, from motor commands and camera images alone? Most tasks are held out.

Canonical URL: https://benchize.com/benchmarks/vals-ai%3Arobo_bench%3Abalance_scale
Last updated: Wed, 21 Oct 2026 18:05:10 GMT

Reproduction and quotation must name Benchize, link to the canonical HTML URL, and retain individual source citations.

Conquered

Can a model control a robot arm and a humanoid in a real-time physics simulation, from motor commands and camera images alone? Most tasks are held out.

Agents & tools · Models tested: 5 · Score publishers: Vals AI

## Model results

| # | Model | Score | ± | Cost | Latency |  |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | [GPT-6.1 Sol](https://benchize.com/models/gpt-6-1-sol-max.md)Max | 100% | SE ±0 pp | $0.2 | 46.54source duration |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 100 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.2038 \| USD \| per_test \| Lower is better \| \| `latency` \| 46.54 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness codex maxOutputTokens Not reported provider openai reasoning true Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0 pp sourcePopulation 7 stderr 0 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_robo_bench.DzDVSjnk.json) Source Vals AI Original model label openai/gpt-6.1-sol Source version 1 Observed 2026-10-21T16:46:05.362Z Evidence SHA-256 `79018880e0e64e5ee8ea196976a86b4ef42ed28e4ec2284db5d302d8aa6226a5` Source locator `/tasks/balance_scale/openai~1gpt-6.1-sol` |  |  |  |  |  |  |
| 2 | [Claude Opus 5.5](https://benchize.com/models/claude-opus-5-5-max.md)Max | 95% | SE ±0 pp | $0.43 | 143.14source duration |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 95 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.4253 \| USD \| per_test \| Lower is better \| \| `latency` \| 143.14 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness claude-code maxOutputTokens Not reported provider anthropic reasoning true Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0 pp sourcePopulation 7 stderr 0 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_robo_bench.DzDVSjnk.json) Source Vals AI Original model label anthropic/claude-opus-5-5 Source version 1 Observed 2026-10-21T16:46:05.362Z Evidence SHA-256 `79018880e0e64e5ee8ea196976a86b4ef42ed28e4ec2284db5d302d8aa6226a5` Source locator `/tasks/balance_scale/anthropic~1claude-opus-5-5` |  |  |  |  |  |  |
| 3 | [Claude Fable 5.1](https://benchize.com/models/claude-fable-5-1-max.md)Max | 10% | SE ±0 pp | $0.8 | 126.42source duration |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 10 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.8025 \| USD \| per_test \| Lower is better \| \| `latency` \| 126.42 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness claude-code maxOutputTokens Not reported provider anthropic reasoning true Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0 pp sourcePopulation 7 stderr 0 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_robo_bench.DzDVSjnk.json) Source Vals AI Original model label anthropic/claude-fable-5-1 Source version 1 Observed 2026-10-21T16:46:05.362Z Evidence SHA-256 `79018880e0e64e5ee8ea196976a86b4ef42ed28e4ec2284db5d302d8aa6226a5` Source locator `/tasks/balance_scale/anthropic~1claude-fable-5-1` |  |  |  |  |  |  |
| 4 | [Gemini 4 Argon](https://benchize.com/models/gemini-4-argon-high.md)High | 0% | SE ±0 pp | $0.52 | 145.52source duration |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 0 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.516261 \| USD \| per_test \| Lower is better \| \| `latency` \| 145.52 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort high harness antigravity maxOutputTokens Not reported provider google reasoning true Reasoning effort high Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0 pp sourcePopulation 7 stderr 0 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_robo_bench.DzDVSjnk.json) Source Vals AI Original model label google/gemini-4-argon Source version 1 Observed 2026-10-21T16:46:05.362Z Evidence SHA-256 `79018880e0e64e5ee8ea196976a86b4ef42ed28e4ec2284db5d302d8aa6226a5` Source locator `/tasks/balance_scale/google~1gemini-4-argon` |  |  |  |  |  |  |
| 4 | [GPT-6 Astra](https://benchize.com/models/gpt-6-astra-max.md)Max | 0% | SE ±0 pp | $1.22 | 38.46source duration |  |
| ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 0 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 1.2184 \| USD \| per_test \| Lower is better \| \| `latency` \| 38.46 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness codex maxOutputTokens Not reported provider openai reasoning true Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0 pp sourcePopulation 7 stderr 0 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_robo_bench.DzDVSjnk.json) Source Vals AI Original model label openai/gpt-6-astra Source version 1 Observed 2026-10-21T16:46:05.362Z Evidence SHA-256 `79018880e0e64e5ee8ea196976a86b4ef42ed28e4ec2284db5d302d8aa6226a5` Source locator `/tasks/balance_scale/openai~1gpt-6-astra` |  |  |  |  |  |  |

---
Cite as: Robo-bench / balance scale — Benchize, https://benchize.com/benchmarks/vals-ai%3Arobo_bench%3Abalance_scale (retrieved 2026-10-21)