MoE | byteiota

Tag: MoE

Apple Silicon chip with streaming data arrows showing MoE expert weights flowing from SSD to 2GB RAM for turbo-fieldfare Gemma 4 26B inference

News

turbo-fieldfare: Gemma 4 26B in 2 GB RAM on Any Mac

turbo-fieldfare streams Gemma 4 26B experts from SSD, fitting a 26B model in 2 GB ...

July 29, 2026

Inkling AI model from Thinking Machines: neural network with adjustable reasoning dial, blue and white color scheme

AI & Development

Inkling by Thinking Machines: Open-Weight AI With a Reasoning Dial

Thinking Machines released Inkling on July 15 — a 975B open-weight model with a unique ...

July 19, 2026

DeepSeek custom inference chip SMIC foundry developer API implications 2026

News

DeepSeek Builds Its Own Inference Chip: What Developers Need to Know

Reuters broke the story: DeepSeek is designing a custom inference chip via SMIC. Here is ...

July 8, 2026

Tencent Hy3 295B Mixture-of-Experts model neural network visualization

AI & Development

Tencent Hy3: 295B MoE Hits SWE-Bench 78 — Free API Ends July 21

Tencent Hy3 is a 295B Apache 2.0 MoE model scoring 78 on SWE-bench Verified. Free ...

July 8, 2026

Mellum2 architecture diagram showing focal model concept with interconnected expert nodes in ByteIota blue and white color scheme

AI & Development

Mellum2: JetBrains Open-Sources a Coding Model Built for Agent Pipelines

JetBrains released Mellum2, an open-source 12B MoE coding model under Apache 2.0. Here is what ...

June 22, 2026

ZAYA1-8B sparse MoE reasoning model trained on AMD MI300X hardware - circuit board with neural network visualization

AI & Development

ZAYA1-8B: Run a Frontier Reasoning Model Without NVIDIA

ZAYA1-8B is an Apache 2.0 sparse MoE reasoning model trained entirely on AMD Instinct MI300X ...

June 11, 2026