Waddle-1 from @theWaddleLabs scores 81% on RoboDojo-RC Tier 1, a real-world robot benchmark.
It is an agent that uses multiple LLMs and shares one context across a task's trials. The other LLMs we tested had a fresh context each trial.
Learn more about Waddle-1 here nitter.gq/theWaddleLabs/status/2…
We're announcing our first robot-use agent, Waddle-1. It controls bimanual robots on unseen tasks zero-shot.
Our agent is the most capable system evaluated by @robocurve, scoring 81 compared to 65 for GPT-6.2 Astra and 59 for Dream-1. It calls LLMs in the cloud to direct lower-level policies that run locally.
Waddle-1 is a first step towards agents that combine the capability of reasoning models and the efficiency of reactive policies. We expect such systems to achieve human-speed manipulation in the near future.
If you would like to deploy robots with our preview API, please tell us what you want to build here: waddlelabs.ai/api-access
1
4
362
The bitter lesson hasn’t reached our wallets yet
Replying to @robocurve
GPT-6.2 Astra (ultrafast) is 70x more expensive than Dream-1, the next highest scoring model.
Robocurve PBC retweeted
One of the reasons why we are incentivizing open-source benchmark creation is because we expect RoboDojo-RC Tier 1 will be saturated by LLMs early next year.
More details about our call for benchmarks here: bounty.robocurve.org.
ARIA is also requesting for benchmark creation here: aria.org.uk/media/fx3f03up/r…
Replying to @robocurve
GPT-6.2 Astra is slightly above the LLM trend on RoboDojo-RC Tier 1. The trend expects LLMs to score >90% by Feb 2027.
Robocurve PBC retweeted
Stand-up bottles was an impossible task for LLMs. Astra is the first to solve it.
Replying to @robocurve
GPT-6.2 Astra is the first LLM we've tested that can stand all three bottles upright. Dream-1 achieved this in October but is a robot model.
Robocurve PBC retweeted
Even though GPT-6.2 Astra is the SOTA robotics model, it is 70x more expensive than the second best model when on ultrafast.
Replying to @robocurve
GPT-6.2 Astra (ultrafast) is 70x more expensive than Dream-1, the next highest scoring model.
GPT-6.2 Astra is the highest scoring model on RoboDojo-RC Tier 1, a benchmark of real robot tasks. It beats Dream-1, the best robotics model we've tested.
Explore all 120 GPT-6.2 Astra runs, with transcripts and videos:
robocurve.org/all-runs/?lead…
1
8
1,574








