Which model benchmarks are prioritized on this platform?

We evaluate models on MMLU-Pro, HumanEval coding, MATH reasoning, token generation latency (tok/sec), and Indic regional language comprehension.

Do you cover both open weights and proprietary API models?

Yes. We cover closed APIs (like Claude 3.5 Sonnet and GPT-4o) alongside open weights (like Llama 3.3, DeepSeek-R1, and Mistral) with self-hosting guidance.

Pillar Hub4 In-Depth Dispatches Active Coverage

Model Launches

Frontier foundation models, open weights releases, architecture papers, and benchmarks.

The Model Launches hub tracks frontier artificial intelligence weights from OpenAI, Anthropic, Google DeepMind, Meta, Mistral, and DeepSeek. We report beyond marketing claims, evaluating context window retention, token throughput, quantization overhead, and real-world reasoning scores across standardized technical benchmarks.

Lead Story in Model Launches

Editorially Featured

Model Launches

DeepSeek-V3 Architecture Deep Dive: Multi-Head Latent Attention and DualPipe Parallelism

How DeepSeek-V3 achieved 671B parameters with only 37B activated per token, training on just 2,048 H800 GPUs for under $6 million.

AI Latest News Editorial Team•October 27, 2026

9 min read

Model Launches

Lead Story in Model Launches

DeepSeek-V3 Architecture Deep Dive: Multi-Head Latent Attention and DualPipe Parallelism

Latest Dispatches (3)

Llama 3.3 70B Benchmark: Matches Llama 3.1 405B Performance at One-Fifth the VRAM

Mistral Releases Pixtral 12B: Open-Weights Vision-Language Model Tested on Document OCR

Anthropic Upgrades Claude 3.5 Sonnet: Computer Use API and Agentic Tooling Evaluated

Frequently Asked Questions about Model Launches

Q:Which model benchmarks are prioritized on this platform?

Q:Do you cover both open weights and proprietary API models?