How to Run Muse Spark Open: Ollama, GGUF, API

How to Run Muse Spark Open: Ollama, GGUF, API

AI at Meta: introducing Muse Spark Open (October 15, 2026) Muse Spark Open in 30 Seconds Meta released the weights of Muse Spark Open on October 15, 2026: 816B total parameters, 42B active, 1M-token context, text, image, video and audio input. Local: the smallest usable GGUF needs about 180 GB of combined RAM and VRAM. Unsloth recommends the 2-bit quant (250 GB+) as the starting point. Ollama: only a cloud tag exists, muse-spark-open:cloud. It runs on Ollama’s servers, not your machine. API: $1.25 input, $0.15 cached, $4.25 output per million tokens on OpenRouter, Together, Fireworks and others. Artificial Analysis scores it 39, the top US open model, behind GLM 5.3 (45) and Kimi K3 (44). “Open weights” and “runs on your laptop” are two different promises, and Muse Spark Open only makes the first one. The weights are a free download on Hugging Face, but even the most aggressive quantization needs a machine with close to 200 GB of memory. For most people the real choice is between an API key, the Ollama cloud tag, or a serious workstation. ...

October 18, 2026 ·  8 min ·  1522 words
Coding Agent Harness: Same Model, 2x the Cost

Coding Agent Harness: Same Model, 2x the Cost

Everyone compares coding agents by asking which model is better. Three independent studies published in the first half of September suggest that is the wrong question. The bigger lever is the harness, and accepting the default can double your bill for the same result, or worse. All three landed within 10 days: HarnessTax from UC Berkeley and Arena (September 16), a 43-page arXiv paper (September 17), and an independent benchmark running nine harnesses against a local model on a laptop (September 10). They agree with each other. ...

September 20, 2026 ·  11 min ·  2303 words
How to Use Claude Code with Ollama, Kimi and GLM

How to Use Claude Code with Ollama, Kimi and GLM

Running Claude Code against a non-Claude model works, and you do it on a path the providers document themselves rather than through a hack. Ollama, Moonshot and Z.ai all publish an Anthropic-compatible endpoint. Setup comes down to three environment variables. This guide covers the steps for all three providers, how to make the configuration stick, and the errors people hit most 🔌 The Short Answer Three providers, three addresses Provider ANTHROPIC_BASE_URL Ollama (local) http://localhost:11434 Moonshot / Kimi https://api.moonshot.ai/anthropic Z.ai / GLM https://api.z.ai/api/anthropic MiniMax https://api.minimax.io/anthropic Alibaba / Qwen depends on the plan Add ANTHROPIC_AUTH_TOKEN and a model name and that’s the whole setup. ...

August 8, 2026 ·  9 min ·  1728 words
Kimi K3: The AI That Designed a Chip in 48 Hours

Kimi K3: The AI That Designed a Chip in 48 Hours

TL;DR: Kimi K3 in 30 Seconds Kimi K3 is the new AI model Chinese company Moonshot AI announced on July 16, 2026; at 2.8 trillion parameters, it is the largest open source model released so far. In a single 48-hour run it designed a working chip, finished an astrophysics study in 2 hours instead of 2 weeks, and cut a teaser video from 56 raw clips. It reads 1 million tokens at once: roughly like reading the entire Lord of the Rings trilogy in one sitting and remembering all of it. On coding it plays in the same league as the strongest Claude and ChatGPT models, and beats them on some tests. It falls behind on hard knowledge questions. You can try it in the Kimi app and on kimi.com. The model files go public on July 27, 2026. Ask an AI to design a chip for you and what happens? It probably writes you a nice article about how chips are designed. ...

July 17, 2026 ·  10 min ·  2049 words