System One
PulseAugur coverage of System One — every cluster mentioning System One across labs, papers, and developer communities, ranked by signal.
- developed by TypeSafe.ai 95%
- used by Clairic 95%
- developed by Clairic 95%
- instance of TypeSafe.ai 90%
- developed Decisions API 90%
- developed by Decisions API 90%
- competes with TypeSafe.ai 80%
- affiliated with TypeSafe.ai 70%
- used by TypeSafe.ai 70%
- competes with GPT-6 Luna 70%
- affiliated with Decisions API 70%
- instance of Decisions API 70%
23 day(s) with sentiment data
Emergence of diverse System One model architectures beyond Jev
The recent introduction of open-source System One models like Laya, Lev, and CLM, which offer distinct architectural approaches (concatenation, multiple-choice, separate encoding), indicates a growing trend towards specialized System One implementations. This diversification suggests that the ecosystem is moving beyond a single dominant model like Jev, exploring different methods to process states and options for decision-making.
System One decision models to gain traction in specific AI agent applications
The benchmarking study comparing System One decision models against classifiers and LLMs indicates their effectiveness varies by task. As System One models generally outperformed zero-shot classifiers on workflow and intent tasks without labels, it's likely they will be increasingly adopted for AI agent applications requiring such decision gates, especially where explicit labeling is not feasible or cost-effective.
System One's Jev model to be integrated into enterprise workflow automation platforms
Glean's Jev model, described as a zero-shot classifier for typed decisions, has seen significant adoption in open-source tools for information filtering and security. Given its performance benefits over LLM-based routers in cost and latency for query classification, it is plausible that enterprise workflow automation platforms will begin integrating Jev to enhance their decision-making capabilities, potentially offering it as a managed service.
-
Small judgment model Jev shows high accuracy but limited impact in coding pipeline
A study involving 5,300 calls to a small judgment model named Jev, developed by TypeSafe, revealed that while the model was highly accurate, its outputs were often not acted upon in a multi-agent coding pipeline. Integr…
-
AI Engineer conference highlights scaling, System One models, and ambitious rewrites; Mistral releases open weights
The AI Engineer (AIE) conference is focusing on key themes for 2026 and 2027, including "Scaling without Slop" to manage AI agent efficiency, the adoption of "System One" models for smarter software development, and inc…
-
Open MedJev model offers efficient zero-shot clinical decision support
Researchers have introduced MedJev, an open-source System One model designed for efficient clinical decision support. This model operates in a single forward pass without generating text, making it computationally effic…
-
Cloudflare launches Clairic extraction models, outperforming GLM-5.3
Cloudflare has introduced Clairic, a new class of extraction models designed for efficiency. These models perform a single forward pass to extract, compute, estimate, and forecast typed values, providing confidence scor…
-
vLLM Semantic Router releases d3-ultra, a 27B multimodal decision model
vLLM Semantic Router has released d3-ultra, a 27 billion parameter multimodal decision model. This model is designed to answer typed questions with probabilities rather than generating text, and it processes input once …
-
Jev-style LLMs show repeatable decisions but not bias resistance
A new research paper published on arXiv investigates selection bias in large language models (LLMs), specifically focusing on Jev-style System One models. The study found that while Jev-style models exhibit repeatable d…
-
Cheap "System One" models rival frontier LLMs in content moderation accuracy
A new research paper from arXiv explores the effectiveness of "System One" models for content moderation, finding them to be accuracy-competitive with frontier LLMs like GPT 5.x, Claude, Gemini-3.*, and Grok, while bein…
-
System One models offer faster, cheaper AI decisions than LLMs
A new class of AI models, termed "System One" models, are designed for fast, structured decisions rather than text generation. In a comparison, a System One model named Jev demonstrated significantly lower latency and c…
-
TypeSafe AI launches Jev, a decision-focused AI model
TypeSafe AI has launched Jev, a new AI model designed for decision-making rather than text generation. Unlike traditional LLMs that produce sequential tokens, Jev uses parallel sampling to provide structured, typed deci…
-
AI research tackles model efficiency and data noise
Two new research papers propose methods to improve the performance and robustness of AI models. The first, EditJev, introduces a system for dynamically routing image editing requests to the most efficient model, signifi…
-
Iris framework enhances multimodal AI decision-making with reusable visual candidates
Researchers have developed Iris, a new framework designed to improve typed decision-making for AI agents, particularly in multimodal settings where candidates are visual. Iris allows for the independent encoding of imag…
-
System One models not better rerankers than fine-tuned alternatives, study finds
A new research paper published on arXiv suggests that "System One" models, also known as decision models, are not superior rerankers compared to models specifically fine-tuned for reranking tasks. The study, which teste…
-
Cloudflare launches Clairic extraction model on Workers AI
Cloudflare has introduced Clairic, a System One extraction model, available in preview on its Workers AI platform. This model is designed to extract specific, typed data fields such as numbers, dates, durations, and lis…
-
VX-One neural network offers faster, more accurate perimeter triage
A new research paper introduces VX-One, a small multi-head neural network designed for real-time perimeter triage. This model outputs typed decision primitives, such as calibrated gates and closed-set choices, aiming to…
-
AI developments span LLM tooling, RL algorithms, panorama generation, and testing frameworks
The llama.cpp project has released version b11781, introducing a POST /tts endpoint for text-to-speech models and enhancing audio streaming capabilities. Separately, a new research paper proposes a model-free continuous…
-
Cloudflare launches Clairic extraction model on Workers AI
Cloudflare has introduced Clairic, a new System One extraction model designed for its Workers AI platform. This model can return typed values along with confidence scores in a single pass, offering an efficient way to p…
-
Cloudflare launches Clairic for rapid data extraction on Workers AI
Cloudflare has introduced Clairic, a new model designed for structured data extraction from text and images. Unlike traditional LLMs, Clairic aims to pull specific information like totals and dates into JSON format in a…
-
Cloudflare unveils new AI extraction models; Apple faces ongoing patent dispute
Cloudflare has introduced Clairic, a new class of extraction models within its System One framework. This development aims to enhance the capabilities of AI in processing and extracting information. Separately, a patent…
-
Cloudflare unveils Clairic extraction model for faster, cheaper data processing
Cloudflare has introduced Clairic, a new type of extraction model within its System One framework. Clairic is designed to extract, compute, estimate, and forecast values from unstructured text with greater accuracy, spe…
-
System One decision models show promise in legal, clinical, and multimodal AI tasks
Researchers are exploring the application of "System One" decision models, which are derived from large language models but focus on outputting calibrated probabilities rather than generating text, for specialized tasks…