Open weights
MiniMax H3
Omni-modal video generation with native stereo audio.
Model index
Generative AI model briefings
Only published profiles are counted. More models appear after source review and final editorial approval.
Reviewed briefing
GPT-5.6 Model Tiers and API Pricing
A source-based guide to choosing between GPT-5.6 Sol, Terra, and Luna, with current API token prices, reasoning options, caching rules, and practical workload boundaries.
Reviewed briefing
Qwen3.8-27B
A documentation-based profile of Qwen3.8-27B, including its native vision-language capabilities, 262K context, thinking controls, supported runtimes, Apache 2.0 license, and official FP8 variant.
Reviewed briefing
UI-Mate-27B
A source-based profile of Tencent’s 27B open-weight GUI agent, including its screenshot-driven action loop, official vLLM deployment path, runtime requirements, and computer-use safety boundaries.
Reviewed briefing
Gemma 4 12B IT
A source-based profile of Google DeepMind’s Gemma 4 12B instruction-tuned model, covering its encoder-free multimodal design, 256K context, Transformers run path, and documented evidence boundaries.
Reviewed briefing
Qwen3.8-Flash-Next
A source-based profile of Qwen’s experimental Flash-Next architecture, including its sparse-attention design, 125B/6B-active model structure, long-context claims, supported runtimes, and deployment boundaries.
Reviewed briefing
Tencent Hy4 Preview
A source-based profile of Tencent’s Hy4 Preview MoE model, covering its 770B/49B-active architecture, one-million-token context, official vLLM and SGLang recipes, FP8 path, and documented limitations.
Reviewed briefing
NVIDIA Cosmos3 Super Image-to-Video 4-Step
An evidence-based guide to NVIDIA?s 64B distilled image-to-video model, its four-step inference paths, supported runtimes, and official multi-GPU hardware boundary.
Reviewed briefing
NVIDIA Nemotron 3 Nano Omni
A source-based overview of Nemotron 3 Nano Omni, including its multimodal architecture, official precision variants, minimum GPU guidance, and supported inference runtimes.