MachinePulseAI Releases Karpo-Decision-31B, Top Open Source Model Beating Google Gemini
MachinePulseAI has launched Karpo-Decision-31B, a 31-billion parameter open-source AI model that achieved the top score of 73.9 on the Decision Index, tying for first place among 300 competitors. Built on the Gemma 4 architecture, the model utilizes a unique readout head to generate zero text outputs, focusing solely on decision probabilities. Benchmarks show it outperforms Google's Gemini Decision 1 in agent call scenarios and offers significantly lower latency (95 ms) and cost compared to Jev 1.14, marking a major advancement in efficient, autonomous decision-making systems.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Reporting timeline
Karpo-Decision-31B Ranks First Among 300 Open Source Decision Models
A new open-source decision model named Karpo-Decision-31B has been launched and reportedly achieved the top ranking out of 300 competing models. The model distinguishes itself by swapping traditional decoding methods for a different approach, though the specific alternative mechanism is truncated in the source text. This announcement comes from the X account @rohanpaul_ai, which highlights the model's superior performance in its category. The release represents a significant development in the open-source AI landscape, suggesting a breakthrough in decision-making capabilities compared to existing alternatives.
Read sourceMachinePulseAI Releases Karpo-Decision-31B Model with Top Decision Index Score
MachinePulseAI has announced the release of a new artificial intelligence model named Karpo-Decision-31B. The model features 31 billion parameters and is distinguished by its ability to generate zero text, focusing instead on decision-making capabilities. According to the source post from DataChaz, this model achieved the top score on the Decision Index, with a specific metric cited as 73.9. The announcement highlights the model's performance in a specialized domain without generating standard conversational or descriptive text outputs. This development represents a specific product release within the AI sector, emphasizing a unique architectural approach to decision indices.
Read sourceKarpo-Decision-31B Beats Google Gemini on Decision Index with Open Weights
MachinePulseAI has released Karpo-Decision-31B, an AI model built on the Gemma 4 31B architecture that achieved a top score of 73.9 on the Decision Index, tying for first place among 300 evaluated models. The model outperforms Google's own Gemini Decision 1 in critical agent call scenarios, specifically recording 90.4 versus 79.9 on When2Call and 90.0 versus 86.2 on Mind2Web. Designed for efficiency, Karpo-Decision-31B operates via a single forward pass to generate probabilities for every option, allowing it to read a page once while answering multiple questions. Benchmarks indicate a median latency of approximately 95 milliseconds on a single RTX PRO 6000 GPU, making it roughly twice as fast as Jev 1.14, and estimated to be 20% cheaper per 1,000 decisions according to JevBench. The weights are open-source, enabling developers to build personal agents or search systems where the model can autonomously act when confident or request human input otherwise.
Read source