Meta's Muse Spark goes open-weight as Muse Spark Open, an 816B-parameter model Meta says is the most capable from the US

Meta released the weights for Muse Spark Open on Thursday and calls it "the most capable US open-weight model to date," "Built for coding and agentic work." It has 816 billion total parameters, 42 billion active per token, and a 1-million-token context window. Meta says it runs on a desktop made of two Apple M5 Ultra Mac Studios, where its demo shipped bug fixes and new features to Ice Cubes, an open-source iOS app. Weights are on Hugging Face under a license with "limited commercial terms," and seven API providers host it today. The "US" in Meta's claim is doing work: on the coding rows of Meta's own benchmark table, Muse Spark Open scores 76.4 on Terminal-Bench 2.1 and 60 on DeepSWE 1.1, behind the Chinese open models Kimi K3 (85.0 and 68) and GLM 5.3 (83.9 and 67).

GitHub Copilot drops Claude Opus 4.8, Opus 5 and GPT-5.6 Terra on November 16

GitHub Copilot drops Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra on November 16, pointing users to Claude Opus 5.5 or GPT-6.1 Sol and GPT-6 Luna instead. Enterprise and Business admins get the replacements automatically under default model enablement, unless turned off; everyone else must update workflows before the date. One exception: individual subscribers on annual plans keep Opus 4.8 past the deadline. Nothing needs removing after the date.

Anthropic's Code Review leaves research preview, and reviews now take 6 to 10 minutes, not about 20

Code Review, the Claude Code feature that reviews pull requests automatically, is out of research preview for Claude Team and Enterprise plans, Anthropic said Thursday. Most reviews now finish "about 3x faster (~20 minutes down to 6-10 minutes)" than in the preview, and admins pick Quick, Standard or Deep review for each repository. Eligible Enterprise organizations can claim at least $10,000 in Code Review credits by January 15, 2027; the credits expire 30 days after they're claimed.

Claude Opus 5.5's fast mode leaves research preview, up to 2.5x faster at double the price

Fast mode for Claude Opus 5.5 is out of research preview, Anthropic's platform notes say: no access request or beta header now, and a default rate limit in the Start, Build and Scale tiers. It runs the same model at up to 2.5x higher output tokens per second, at $8/$40 per million tokens, double Opus 5.5's standard $4/$20. It's now on Claude in Microsoft Foundry. Fast mode for Opus 5 and 4.8 stays in preview.

Anthropic adds a Mononoke credit to its Scion post after a builder's complaint

Rain, an engineer who helped build Mononoke, Meta's source-control server, said Anthropic's post on Scion, its in-house git replacement, copied Mononoke's design without credit. Anthropic has added a footnote: Mononoke "pioneered many of the techniques we use in Scion." Nick G., the post's author, announced it in a Hacker News reply, saying the post credited Sapling, Meta's name for the whole system. Rain calls the credit sufficient but stands by the complaint, and says they privately heard the copying was intentional, with "overlap in personnel."

Also worth your time

Know someone who'd want this in their inbox? Forward it — that's how this grows. And if we got something wrong, or you think we buried the real story today, hit reply. A person reads every one.

The New Way is written with AI. It gathers the day's stories, checks them against their sources and drafts every summary. A person decides what runs and reviews every issue before we hit send.

Meta's Muse Spark goes open-weight as Muse Spark Open, an 816B-parameter model Meta says is the most capable from the US

What changed this week