Singularity Compute, which runs on @SingularityNET's tech, has released Super Decider, a model that anyone can download and use for free under an open license. Instead of writing text, it pic
Singularity Compute, which runs on @SingularityNET's tech, has released Super Decider, a model that anyone can download and use for free under an open license. Instead of writing text, it picks from options a developer sets, such as which team should handle a support ticket, and gives each option a probability in about 0.3 seconds.
The model is live on ASI:Cloud (@CUDOS_). In the company's own testing, it scored 85.6 on NVIDIA's When2Call benchmark, ahead of TypeSafe's Jev and Cloudflare's Clef, but 1.6 points behind Clef on a tool-selection test. The figures come from the developer and have not been independently verified.
According to the project's GitHub repository, Super Decider answers typed questions about a state, covering choices, yes/no questions and scores, each with a calibrated probability distribution returned in a single call. The repository describes it as a fine-tune of Google's DiffusionGemma 26B-A4B, a discrete-diffusion mixture-of-experts model with 4 billion active parameters. It reads every answer slot in one denoising step and runs on a single NVIDIA L40S GPU.
The same page says the API follows the Jev / TypeSafe "System One" request format and is also compatible with OpenAI-style chat completions. The code is released under the Apache-2.0 licence. The developers state that no benchmark test items were used for training.
For developers, the appeal is a structured output rather than free-form text. A classifier-style model that reports its own confidence can be used to route support tickets, triage requests or gate automated actions, with low-confidence cases passed to a human.
When2Call is an NVIDIA benchmark. Its peer-reviewed paper says it evaluates tool-calling decision-making: when to generate a tool call, when to ask follow-up questions and when to admit a question cannot be answered with the tools provided. That focus on the decision, rather than the accuracy of the call itself, is why it suits a model built to choose between options.
The result sits within a broader push by the ASI Alliance into AI infrastructure. Singularity Compute describes itself as part of SingularityNET and the ASI Alliance, and delivers inference through the ASI:Cloud platform. The alliance links Fetch.ai, SingularityNET and CUDOS under a shared token, according to Crypto.com Academy.
Because the benchmark results are self-reported and the comparison with Clef on tool selection is narrowly negative, developers may want to test the model on their own workloads before adopting it.
Sources
Super Decider repository, Singularity Compute (GitHub)When2Call: When (not) to Call Tools, ACL AnthologyWhat Is the Artificial Superintelligence Alliance?, Crypto.com
