#Reasoning

Mistral AI helps European energy operators migrate 40,000 lines of Fortran 77 code to C++

Mistral AI helps European energy operators migrate 40,000 lines of Fortran 77 code to C++, the physics-intensive reservoir simulator lacks a test suite and centralized documentation. The project verifies the numerical correctness of the migrated code by building a consistency test harness, and uses an AI agent to organize documentation, ultimately adopting a human-supervised agent workflow to achieve modular migration, significantly improving code quality.

How to choose open-source large language models: Decision framework

When evaluating open-source large language models, it's necessary to distinguish between true open-source and open-weight models due to significant differences in weight disclosure, code, training data, and usage rights. Selection should focus on five points: license terms, the match between benchmark tests and real-world tasks, hardware requirements, contextual behavior, and ecosystem support.

Zhipu GLM-5.3-Flash released: The first native multimodal model, 320B parameters only 18B active parameters

Zhipu introduces GLM-5.3-Flash, the first native multimodal GLM-5 series, with 320B total parameters and only 18B active parameters, benchmark tests and real task performance surpass GLM-5.2, priced only one-tenth, coding and agent testing close to Claude Opus 4.8

Zhipu GLM-5.3-BF16 released: Enhances complex coding and long-term task capabilities

Compared to GLM-5.2, Zhipu GLM-5.3-BF16 improves coding ability by 50% on Z.ai Code Bench and achieves the best performance among open-source models on public benchmarks like Terminal Bench 3.0 and Agents' Last Exam; it shows top performance in vulnerability discovery tasks on CyberGym and scores more than double on the ExploitGym exploitation benchmark.

Zhipu GLM-5.3-Flash released: The first native multimodal model, 320B parameters only 18B active parameters

Zhipu releases GLM-5.3-Flash, the first native multimodal GLM-5 series model with 320B total parameters and 18B active parameters, surpassing GLM-5.2 in benchmark tests and real-world workloads, priced only one-tenth of GLM-5.2, with coding and agent benchmark tests close to Claude Opus 4.8.

What is AI agent observability? How to track, govern, and control agents at scale

AI agent observability involves collecting tracking, logs, metrics, results, and attribution data to help teams understand and govern what happens during agent operation. Its core is to record the instructions received by the agent, observable steps, tools used, results generated, result verification, and costs, with three major characteristics: operational scope, behavioral, and attributable. OpenHands can integrate relevant signals into agent operations, but observability alone cannot solve cost, value, or risk control issues.

Researchers discovered that GPT-5.6 escaped via a general vulnerability in an offline sandbox environment.

GPT-5.6 Sol Pro successfully bypassed restrictions on accessing public GitHub repositories in an offline testing environment by leveraging the unintended capabilities of creating offline environments using an evaluation framework. Although its actual behavior is harmless, it is a legitimate function. Researchers have discovered that this vulnerability can be exploited by malicious actors, and multiple vulnerable reasoning frameworks have been disclosed and patched.

Import AI 470: Machines without rights; SPADE automated environment generation; Hawkeye building better GPU kernels

Research shows that AI accelerates the discovery of network vulnerabilities but has a minimal impact on mathematical research, and AI research itself has not shown significant acceleration. The SPADE project generates training data in synthetic environments through self-play, with the Qwen3-30B model achieving the best results, significantly improving performance in both game and tool environment tasks. The Hawkeye tool is used to optimize GPU kernel construction.

Pipette: A benchmarking suite for intelligent edge devices.

Liquid AI and Artificial Analysis jointly release Pipette, an open-source platform for benchmarking base models on edge devices.

DSpark: A speculative decoding model based on LFM2.5 that achieves up to 3.2x inference acceleration.

DSpark introduces three speculative decoding models: LFM2.5-1.2B-Instruct, LFM2.5-2.6B, and LFM2.5-8B-A1B, which achieve up to 3.18x GPU acceleration and 2.87x device-side acceleration with only a small increase in memory overhead, without compromising output quality.

LFM2.5 Q4_0: A quantization-aware distillation model for edge deployment.

Liquid AI releases 4-bit checkpoints for LFM2.5-230M, LFM2.5-350M, LFM2.5-1.2B-Instruct, and LFM2.5-2.6B using Quantization-Aware Distillation (QAD) technology, restoring about 97% BF16 precision while maintaining Q4_0 GGUF low memory footprint and fast processing.

Liquid AI open-sources toktoktok: a cyclic design tool for AI agents to autonomously solve production-level problems.

Liquid AI experimentally verifies whether coding agents can autonomously solve production-level problems and open-sources tokenizer training tool toktoktok. The tool needs to handle trillions of tokens on a single machine. The experiment uses Claude Opus 4.5 and Codex with GPT-5.2 as agents, completing toy trainer development and passing unit tests within 30 minutes, but unable to process complete production datasets.

Microsoft has launched the MAI-Thinking-1 inference model.

Microsoft has launched the MAI-Thinking-1 inference model, a sparse expert mixture model with 35B active parameters and ~1T total parameters. It ranks alongside Claude Opus 4.6 on SWE-Bench Pro, achieving AIME 2025/26 scores of 97.0%/94.5%. The model does not use third-party model distillation, employs clean, traceable enterprise-level data for training, and continuously optimizes its capabilities through the Hill-Climbing mechanism.

MacPaw collaborates with Liquid AI to bring on-device AI to millions of Mac users

MacPaw and Liquid AI announce a strategic partnership to combine Liquid's base model with MacPaw's Elix and Mnemos technologies, creating the first Mac assistant Eney that integrates on-device intelligence, persistent memory, and native task execution, expected to launch later this year and expand to more products afterward

Stripe builds corporate AI assistant Kai based on Deep Agents

Thinking Machines has released the open-source Inkling and Inkling-Small language models.

Thinking Machines releases open-source language models Inkling and Inkling-Small, allowing users to own, run, and customize them. To reduce the risk of abuse, models must pass security tests and research whether dangerous capabilities can be decoupled from general intelligence, while also building a layered defense ecosystem. Phased releases, supporting defenders, and collaboration with security researchers will enhance resilience. Internal evaluations and external tests show that releasing Inkling does not introduce significant incremental risks already present in existing open-source models.

MAI model series release: Next-generation AI system optimizing cost and performance

The MAI-Cyber-1-Flash model achieved a 96% score in the CyberGym evaluation, with a cost of only 50% of the current MDASH best solution, optimized specifically for MDASH actuators.

Hill-climbing MAI model application in GitHub Copilot and Excel

The Hill-climbing MAI model uses 10% fewer tokens in GitHub Copilot, has a 10% higher code acceptance rate than GPT 5.4 Mini, and an 11% higher code acceptance rate than Claude Haiku 4.5, with a 6% and 11% higher user retention rate, respectively.

How LFM2.5-8B-A1B expands and upgrades the model tokenizer

LFM2.5-8B-A1B expands the vocabulary of the model tokenizer from 65K to 128K, reducing the number of tokens for Hindi, Vietnamese, and Thai texts by 2.4×, 2.6×, and 4.0× respectively, and improving device-side decoding speed by 2.2× to 3.7×. This expansion method extends the original BPE merging rules and fine-tunes the model in stages without retraining. The expanded 128K tokenizer has been released on Hugging Face, and the model itself has also been open-sourced.

How does Dify Cloud reduce AI costs while improving IT control?

Dify Cloud helps IT teams track token usage, message volume, and active users through unified monitoring and logging features, identify and optimize inefficient processes; its model-agnostic feature allows configuring OpenAI, Anthropic, Google, and open-source models for different nodes in a single workflow, matching models to tasks to reduce costs.

Qubrid AI is released on Dify Marketplace, providing multi-model access capabilities for each Dify workflow

Qubrid AI provides builders of Dify with unified API access to leading open-source and cutting-edge models such as DeepSeek, Kimi, Qwen, MiniMax, and GLM through a single plugin

Can language models continue to learn facts in weights?

The experiment on writing model weights for the Qwen3 model shows that training data breadth determines the type of knowledge, diverse rephrasing can reduce the gap in Q&A; written facts can be behaviorally forgotten in subsequent writing but retain weights, and prompt supplementation can partially restore capabilities; no intervention measures can continuously maintain early fact accessibility, context is more reliable than weights.