Tag
Levanto Labs introduces Sage2, a streaming decision model that computes each token of a conversation only once, reusing cached context so each new message adds only its own token cost. It delivers typed answers with calibrated probabilities and claims 3x faster decisions on progressive context, plus extraction that is 166x cheaper than Opus 5.5.
Nace.AI released Drex Max v1, an open-weight 27B multimodal decision model with a 255-way readout head and 262K-token context that returns probability distributions over options in a single forward pass, scoring 67.52 on the public Decision Index 0.4 and running locally via llama.cpp, Ollama, and MLX on Apple Silicon.
Articul8 AI, a company spun out of an Intel team, is launching Bivia, a multimodal AI decision model designed to complete entire production workflows across industries like semiconductors and energy, rather than just answering individual questions.
Nace 发布 Drex Max 1.1，这是一个在分布外泛化上显著提升的多模态决策模型，拥有 262K 上下文窗口，并在 Decision Index 上取得 70.8 的成绩。
Liquid AI released d1-2B, a 1.92B-parameter decision model for text, images, and video built on LFM3-VL-2B that returns calibrated typed answers in a single forward pass with no token generation, achieving millisecond-scale inference and leading open d1 benchmarks.
Mica v0.5 是一个 4B 参数的决策模型，在单张消费级 GPU 上实现极速推理（256 token 仅 19.8 ms），并在 JevBench Hard 和 JDI 上取得 4B 级别最佳成绩，采用 Apache-2.0 开源协议发布。
Nace releases Drex Max 2.0, a 27B multimodal decision model for computer-use agents that scores options in a single forward pass with millisecond latency, priced at $0.06 per million input tokens via the Nace API.
Mass General Brigham researchers present Clinical System One (CS1-4B), a 4B-parameter decision model that performs fast, typed judgments over clinical records, outperforming open decision models up to 27B parameters on 13 of 14 tasks while running on a single desktop GPU. Paired with a generalist LLM agent, it enables longitudinal chart review and clinical trial screening with 690-fold reduced token usage.
Mouse-1 是一个面向 agent harness 的轻量决策模型，能处理 agent 执行过程中的大量小决策，宣称比 GPT-6 Sol 便宜 163 倍、快 65 倍，可与现有技术栈并行接入。
qev-0.8b is a 0.8B-parameter local decision model that takes a single forward pass to return calibrated probabilities over any input, ranking #1 of 27 similar-sized models on Benchmark Heaven (77.1%) at 3.5 ms p50 latency while running in 1.3 GB on a laptop. It ships in CUDA, MLX and GGUF formats and generalizes to games (Tetris, Pac-Man, 2048, Breakout, Whac-A-Mole) never seen in training.
用户分享了使用 Jev（Qwen 3.8 27B LoRA 决策模型）驱动的贪吃蛇游戏，模型在每步操作时根据棋盘状态文本描述进行决策，延迟约 600ms，运行于 llama.cpp 环境中。
TypeSafe's Jev decision model now supports multimodal input with a new Box question type and assets field, returning structured, schema-safe answers (choice, noul, score, box) with probabilities directly usable in code instead of parsing free-form LLM JSON output.
Jev decision model now supports multimodal input, allowing images in both state and questions, plus a new Box type that returns YOLO-style bounding boxes for object localization. The update also supports parallel question answering for interactive vision applications, with images and Box still in experimental status.
Liquid AI 发布 d1-ultra，一款多模态决策模型，在 112 项评测中超越 Jev 1.14 及所有其他决策模型，输出经校准的概率答案，现已在 Liquid API 上线，价格为每百万输入 token 0.042 美元。
Liquid AI released d1-ultra, a multimodal decision model that processes text and images in a single forward pass to return calibrated probabilities without token generation, claiming top performance on the Artificial Analysis Decision Index and pricing at $0.042 per million input tokens.
作者将 Jev 决策模型集成到开源 agent harness TrueForge 中，实现比让 LLM 做全部决策低 150 倍成本的结构化选择，并结合 Claude Opus 复核低置信度结果，保持同等准确率。
作者介绍了 DecisionLoop，一个用小型决策模型替代 LLM 处理意图识别、工具选择与控制流的 Python 智能体框架，初步测试在工具调用准确率、成本和延迟上优于基于 LangChain 的 LLM 智能体，并寻求架构反馈与基准测试建议。
推文讨论 SubQ Snap 的延迟数据：100K tokens 约 660ms、384K tokens 约 2.5 秒，接近线性扩展而非二次注意力的 15 倍增长，使决策模型能一次处理超长多模态上下文。
SubQ 发布长上下文决策模型 Snap，支持 384K 多模态上下文、每 token 计算成本恒定，并以极低定价（$0.05/百万输入 token，输出免费）冲击企业级 AI 经济性。
Subquadratic 发布 SubQ Snap，号称首个面向企业级真实工作负载的长上下文决策模型，在 Hugging Face Decision Index 上取得 61.6/61.7 的成绩，超过 Jev 的 57.9，且声称每 token 计算量不随上下文增长而增加。