AI Briefing#reasoning-model
The latest AI and developer news about #reasoning-model, with the original source and a short summary.
The latest AI and developer news about #reasoning-model, with the original source and a short summary.
The model supports text, image, and video inputs with adjustable reasoning effort and is accessible via AI SDK, Chat Completions, Responses, and Anthropic Messages APIs.
Cat Jam is free to use during its stealth period and supports function calling and structured outputs for agentic tasks.
The open-weight model achieves 184 tokens/s on an Apple M5 Max and scores 48.09 on LiveClaw, outperforming larger models in agentic tasks.
The model is free for Pro and Enterprise teams with purchased AI Gateway credits but lacks Zero Data Retention support.
Ember-1 is a research preview model built on Kimi K3 that reportedly generates 40% fewer tokens than its base model.
IBM released three Granite 4.2 reasoning LLMs featuring 512K context and agentic RL.
Support for the new inference model Ling 3.0 has been merged into llama.cpp.
Financial-RLVR-10K, a 10,000-sample financial reasoning dataset containing executable Python code, has been released.
Microsoft has released MAI-Thinking-1, a high-performance reasoning model optimized for enterprise tasks.
DeepSeek V4 Flash 0731 achieved 89.0% on ARC-AGI-1.
LLM 0.32 supports reasoning traces, server-side tools, and OpenAI-compatible endpoints.
Reasoning-Medical-27B, a model built on Qwen3.6-27B that maximizes medical reasoning ability, has been released.
XYZ AI Lab has released XYZ-Aquila-mini, a deep search agent model based on Qwen3.6.
The ARC-AGI-3 leaderboard, which measures the adaptive intelligence of AI agents, has been released.
LG AI Research has unveiled 'EXAONE Deep,' an AI model with overwhelming reasoning performance in math, science, and coding.
This paper proposes a diagnostic framework that analyzes, at the token level, the impact of On-Policy Distillation on training reasoning models.
An 800M-parameter model that summarizes coding agents' complex reasoning traces into structured JSON format has been released.
Proposes a risk control framework that optimizes token usage during LLM inference while keeping the error rate below a certain level.
China's Weibo has released a 3B-scale VibeThinker model that has become controversial after posting benchmark scores that surpass ultra-large models.
VibeThinker-3B, a small language model (SLM), has demonstrated frontier-level performance on math and coding benchmarks.
SupraLabs has released a 50M-scale reasoning model that generates a thinking process using a synthetic dataset.
A Thinking mode toggle that lets users control a model's reasoning process has been added to the llama.cpp UI.
Liquid AI has released an on-device 8B-A1B MoE model trained on 38T tokens with enhanced reasoning capabilities.
The claim is that because a reasoning model's reasoning trace and final answer are generated from the same computation, the reasoning trace cannot guarantee the faithfulness of the answer.