Models
Multi-agent AI needs three governance modes, Australia’s AISI argues
Australia's AI Safety Institute maps multi-agent systems into singular, federated and open governance. Here is the control stack builders can apply.
Read more
Topic
Models
Frontier models, comparisons, and how to choose what fits your work.
30 articles
Musthave.AI — AI models, tools, and practical news for builders
Models
Granite 4.2 is local and Apache-licensed. Its context can still break a 16 GB budget
IBM Granite 4.2 adds open 3B, 8B and 30B reasoning models. The practical test is memory and accepted tasks at fixed context lengths.
Read more
Models
Hugging Face says agents became its number-one user. Downloads still do not mean model quality
Hugging Face reports 2.96 million public model repos and rising agent traffic. The download distribution still measures activity, not model quality.
Read more
Models
NVIDIA says Vera Rubin runs 30× more agent work per megawatt. The result still awaits outside review
NVIDIA reports up to 30× more agent throughput per megawatt on Vera Rubin. The benchmark is early, company-run, and awaiting independent review.
Read more
Models
Fable 5’s NanoGPT Speedrun used 811 experiments to close one gap
Prime Intellect ran 153 autonomous optimization attempts. Fable 5 led the NanoGPT Speedrun, but the result measures a model-harness-effort system.
Read more
More stories
Thomson Reuters’ Thomson model cost $40M. The outside test is still missing
NVIDIA says AVO hit 100 on ARC-AGI-3. The private set is still the test
LLMRouter puts 16 model-routing strategies under one test
Nemotron 3.5 Lightning is a 66 GB model with 3B active. Do not confuse compute with memory
Meta Muse Glimmer fits in 17 GB. Day-one backend support is the real test
An AI agent rewrites its safety harness after failure. The held-out test matters
Text-to-SQL memory only matters when a repair helps the next question
Mistral Shieldstral changes policy without retraining. The threshold is still yours
This AutoML benchmark lost 25 points when the clock was enforced
P-Bench shows an AI can run the code and choose the wrong statistical test