
Qwen3 explained: dense models, experts, and thinking modes
The Qwen3 technical report describes dense and mixture-of-experts models, with support for thinking and non-thinking behavior.
AI model news, releases, cost, and comparisons. What shipped, who measured it, and which figures are still vendor claims.

The Qwen3 technical report describes dense and mixture-of-experts models, with support for thinking and non-thinking behavior.

Meta’s Llama 3 report describes a family of foundation models and instruction-tuned variants.

Mistral 7B is the language model described in Mistral’s 2023 paper.

Mixtral 8x7B stores eight experts per layer and runs two of them for each token. Mistral reported 47 billion parameters in reach and 13 billion active.

Gemma 2 is the open-weight model generation described in Google’s technical report.

Microsoft’s Phi-3 report studies compact language models trained with an emphasis on data quality.

OLMo 2 is a language-model family documented by the Allen Institute for AI and collaborators.

SmolLM2 is a compact language-model family from Hugging Face.

Whisper is a speech-recognition system described in OpenAI’s research on weakly supervised audio training.

CLIP, from OpenAI in 2021, learns which caption goes with which image across 400 million pairs. Text can then point at a picture without a new labeled training set for every category.

Sentence-BERT turns a BERT model into sentence vectors you can compare with cosine similarity. The 2019 paper cut a 10,000-sentence similarity search from about 65 hours to about 5 seconds.

BERT stands for Bidirectional Encoder Representations from Transformers.

T5 explores a unified text-to-text approach to language tasks.