Explore models, discover tools and understand the advances that are changing technology.
Verified analysis
Nemotron 3.5 Transcribe combines multilingual transcription with translation, timestamps and speaker labeling. NVIDIA says it can process recordings of up to an hour; reported performance figures come from specific evaluations and do not establish equivalent results across every language or use case.
CrowdStrike says VAPOR PANDA used ARTEX, primarily with DeepSeek, against more than 200 organizations in at least 14 countries. The company confirms post-intrusion activity at nine; the target count is not a count of confirmed victims.
The fifth edition adds two evaluations and updates three others. Artificial Analysis ranks Claude Opus 5.5 first and puts the gap to the highest-scoring open-weights model at 12 points. The published scores have not been independently verified in the available evidence.
Six routes from headline to informed decisions, comparing alternatives and checking their primary sources.
Verified profiles, capabilities and versions.
What they solve and for whom.
Planning, memory and action.
APIs, patterns and evaluation.
Privacy and control without cloud.
The papers, without losing rigor.
Select up to three models to contrast their orientation and modalities.
Generalist frontier model oriented towards complex tasks, use of tools and highly demanding professional work.
Model oriented to code and knowledge work, with controls that allow effort to be adjusted to each task.
Flash model for long-running software engineering, autonomous agents, and complex business flows.
Complex work, coding, and tool-using agents.
Tasks requiring a balance of quality, speed, and cost.
High volume, subagents, and structured extraction.
i Profiles show their verification date and always link to the primary source. Missing data is never inferred.
Models (140) ↗Memory, quantization, context and speed: a practical route to fit the model to the computer and not the other way around.
Attention, useful signal and practical limits.
A guide to interpreting long benchmarks.
What changes when the inference is local.
Less infinite directory; more context about the problem each tool solves.
An editorial reading of how the product focus has changed. It is not a benchmark: it summarizes recurring capabilities in the official announcements analyzed.
Qualitative editorial index · updated SEP 2026
Each route connects a need with evaluation criteria, not with a specific brand.
Privacy, long context and citations as minimum criteria.
View relevant criteria →Understanding the code, tools and automatic verification.
Compare capabilities →Local memory, quantization and control of the environment.
Open local AI guide →Selected modelsChoose between two and three to compare