LiveTamil Nadu's maiden TVK Budget draws industry praise, educationists' ireLiveMartin Spinoff The Varnell Hill Show Premieres Sept. 1 on Paramount+; Netflix Unveils Lizzie Borden SeriesLiveCyberattacks on US water systems reported in at least 7-12 states, possibly Iran-linkedLiveInfantino stays as Fifa president, apologizes for investment plan errorsLiveRussia Preparing Fresh Mobilization Drive, Kyiv and Independent Media Report; Kremlin DeniesLiveIran Says Deal with Oman on Hormuz Shipping Route in Final StagesLiveLIV Golf secures new lead investor as Saudi funding ends after 2026LiveRosie O'Donnell Says She Won't Worry About FCC During Kimmel Guest Hosting StintLiveTamil Nadu's maiden TVK Budget draws industry praise, educationists' ireLiveMartin Spinoff The Varnell Hill Show Premieres Sept. 1 on Paramount+; Netflix Unveils Lizzie Borden SeriesLiveCyberattacks on US water systems reported in at least 7-12 states, possibly Iran-linkedLiveInfantino stays as Fifa president, apologizes for investment plan errorsLiveRussia Preparing Fresh Mobilization Drive, Kyiv and Independent Media Report; Kremlin DeniesLiveIran Says Deal with Oman on Hormuz Shipping Route in Final StagesLiveLIV Golf secures new lead investor as Saudi funding ends after 2026LiveRosie O'Donnell Says She Won't Worry About FCC During Kimmel Guest Hosting Stint
← Back to home
Tech

Meta launches Muse Code AI coding agent and Muse Spark 1.2 model

Meta has released Muse Code, a terminal-based AI coding agent now in beta, alongside Muse Spark 1.2, a new coding-focused AI model. The launch puts Meta in direct competition with Anthropic's Claude Code, OpenAI's Codex and other agentic coding tools that have become central to how many developers build software.

Muse Code is designed to handle complete software engineering tasks across large code repositories: planning changes, writing code and validating results. It installs on macOS or Linux with a one-line command, though users must log in with a Meta account and provide billing details. CEO Mark Zuckerberg announced the release in a post on X.

The agent's main architectural feature is what Meta calls async background agents. Instead of spawning helper agents fresh for each task, Muse Code keeps specialized background agents alive for the entire session, avoiding redundant information gathering. For large jobs, it fans out to parallel sub-agents, each working in an isolated git worktree so the developer's own copy is never touched. Meta said it tested the agent building six features for a game simultaneously with no collisions.

Muse Code also maintains a local event log of every model call, tool run, approval and edit, making the runtime "replay-exact and restart-safe" -- if it crashes mid-task, it resumes from where it stopped without losing work. It ships with bundled skills such as /plan, which turns a task into an approval-gated plan, and /grill, which stress-tests the plan.

Muse Spark 1.2, the model behind the agent, is a coding-focused update to Muse Spark 1.1 with "significantly scaled up" training compute on coding tasks. Meta co-trained it with Muse Code itself, using the agent's trajectories to improve performance inside the tool. The company also used a self-improvement loop: Muse Spark 1.1 generated challenging coding environments, then graded candidate solutions to build a training dataset for the successor.

When the original Muse Spark debuted in April, it scored 77.4 on SWE-Bench Verified, trailing Claude Opus 4.6's 80.8 and Gemini 3.1 Pro's 80.6. As a demonstration, Meta pointed Muse Spark 1.2 at GPU kernel optimization and let it run for more than 1,000 tool calls over up to 24 hours on NVIDIA Hopper hardware. The agent wrote, compiled and profiled code, and Meta says it found "substantial improvements" over baseline implementations.

Muse Spark 1.2 is available through Meta's Model API in two pricing tiers. The standard tier costs $1.25 per million input tokens and $4.25 per million output tokens, with prompts and completions not used for training. A "contributor" tier costs $0.10 per million input tokens and $0.20 per million output tokens -- roughly 12x and 21x cheaper -- but in exchange Meta may use the prompts and completions to train future models. Even the contributor tier requires a payment method on file, and it carries tighter rate limits: 60 requests per minute versus 3,000 on standard.

The launch is notable for what it does not mention: open source. Meta's Llama family of open-weight models had been downloaded roughly 1.2 billion times, but the new Muse models are proprietary, cloud-only and have no downloadable weights. Rivals OpenAI and Google have released their coding CLIs under open-source licenses, while Meta's tool is proprietary like Anthropic's.

This summary was written by AI from the source reports listed above. It may contain errors. Please check the original source before relying on it. Spotted something wrong? Email auditor@trendingwire.news and we will correct or remove it.
Tech30 min ago
Cyberattacks on US water systems reported in at least 7-12 states, possibly Iran-linked
Cyberattacks on U.S. water systems have hit at least a dozen states, according to CBS News, and officials suspect Iran-backed hackers are responsible. Fox News
CBS News, Fox News
Tech2 hr ago
Jeff Dean Leaves Google to Launch AI Startup; Demis Hassabis Steps Down as DeepMind CEO
Google's chief scientist Jeff Dean is leaving the company to launch a startup with other departing Google executives that will use artificial intelligence to ad
TechCrunch, CNBC
Tech14 hr ago
Apple Asks Court to Block OpenAI's Use of Alleged Trade Secrets; OpenAI Fires Back
Apple has asked a federal judge to immediately block OpenAI from continuing to use trade secrets Apple says were stolen, while a lawsuit between the companies i
NDTV, Fortune
Tech1 days ago
India Summons Meta Global Team Over CSAM Lapses, Account Action Issues
The Indian government has summoned Meta's global team to India for meetings on August 5-6 to discuss a range of concerns, including lapses in handling child sex
Hindustan Times, The Hindu
Tech1 days ago
WhatsApp Blocks Some Accounts in India for 24 Hours as Users Report No Warning
Several WhatsApp users in India said they were locked out of the messaging app for 24 hours after their accounts were placed under review, according to reports.
Hindustan Times, The Hindu
Tech1 days ago
Hugging Face CEO Warns China Could Overtake US in AI This Year
Hugging Face CEO Clément Delangue said Chinese AI models could catch up to the United States as soon as this year, warning that China is winning the AI race. He
Times of India, CNBC