Muse Spark 1.2

OpenCode Go is a $10/month subscription tier from OpenCode that gives access to popular open coding models with stable global access, intended primarily for international users. It works like any other OpenCode provider: users sign in to OpenCode Zen, subscribe to Go, copy an API key, run /connect in the TUI selecting The Go lineup page, published 13 hours before 2026-09-02, explicitly lists Muse Spark 1.2 as one of the available models, tagged as "Contributor (limited regions)" — an important regional-restriction caveat for users. The broader Go model list includes Grok 4.6, GLM-5.x variants, GPT 5.6 Luna, Kimi K2.6/K2.7 Code/K3, L

Note.com's AI-Driven Lab piece (dated August 6, 2026) analyzes the August 5, 2026 launch of Meta's terminal-based coding agent "Muse Code" and its underlying model "Muse Spark 1.2," drawing solely on primary sources — Meta's official blog and evaluation methodology report. It covers the resident background agent archit The piece positions Muse Code against Claude Code and Codex as the entry point for evaluating the model, noting Meta as a long-time outsider in AI coding tooling. It frames the model's release alongside Muse Code as Meta's deliberate entry into the coding-agent market rather than relying on third-party CLIs to unlock t

Daily.dev's Aug 11, 2026 write-up details Muse Spark 1.2's launch on August 5, 2026 alongside Muse Code, Meta's first terminal-based AI coding agent, both from Meta Superintelligence Labs. Muse Code installs with a single command and targets full software-engineering workflows (planning, writing, validating code) rathe On coding-specific benchmarks, Muse Spark 1.1 scored 53% on DeepSWE (trailing GPT-5.6 Sol at 73% and Claude Opus 5 at 74%), while Muse Spark 1.2 reaches 60.3% on WeirdML, comparable to GPT-5/GPT 5.4 Mini. The clearest 1.2 gains are agentic: GDPval-AA v2 Elo rose 260 points (1371→1631), Terminal-Bench climbed from 78% t

According to a CNBC report dated August 10, 2026, Meta CEO Mark Zuckerberg announced in an Instagram video that the company would open the weights of its Muse Spark 1.2 model, allowing the public to download and use it. The article notes that "weights refer to the calculations and rules that determine how the AI works The CNBC piece contextualizes the release within Meta's investor narrative: shares were up 2.1% in premarket trade following the announcement, and the article ties the open-weights strategy to a 6,500-word essay by Zuckerberg urging Washington to support open-source AI against Chinese competitors. Creator attribution i

A Ginger Labs post dated August 5, 2026 focuses on the Muse Spark 1.2 Contributor API tier, framing the article around three evaluation questions: whether the model completes workflows reliably in a developer's harness, whether the contributor data-use bargain fits the intended data, and whether a system can survive a The post surfaces operational and portability risks specific to Muse Spark 1.2: the 1-million-token context window is useful but easy to misuse, account-level operational limits remain poorly documented, and the tight coupling between Muse Spark 1.2 and the Muse Code harness creates a portability risk for teams relying

Digital Applied's August 6, 2026 launch guide documents that Meta released Muse Spark 1.2 alongside Muse Code (in beta) on August 5, 2026, framing the model and its terminal coding agent as a paired product rather than a standalone model release. It reports the Meta Model API list pricing of $1.25 input and $4.25 outpu The guide highlights that Meta shipped its own command-line agent (Muse Code) co-trained with Muse Spark 1.2, drawing a parallel to Anthropic's Claude Code and OpenAI's Codex. It frames pricing comparison as secondary to the strategic shift of bundling harness and model, and notes the piece draws on the Meta research b

Yahoo Finance's August 5, 2026 technology report covers Meta's debut of Muse Spark 1.2 and Muse Code, its first coding agent from Meta Superintelligence Labs (MSL). The article describes Muse Spark 1.2 as a coding-centric update to the Muse Spark line, with Muse Code as the coding harness that runs the model, and posit The same piece situates the launch against Meta's recent AI investment cycle, noting CEO Mark Zuckerberg's $14.3 billion acquihire of Scale AI CEO Alexandr Wang (now Meta's chief AI officer) and prior Muse Spark, Spark 1.1, Muse Image, and Muse Video releases. It reports Meta compared Muse Spark 1.2 against Opus 5, GPT

Artificial Analysis published a benchmark analysis of Meta's Muse Spark 1.2 on August 5, 2026, scoring it 54 on the Artificial Analysis Intelligence Index and calling it Meta's third release in four months. The article notes Muse Spark 1.2 closes prior gaps on agentic knowledge work: its GDPval-AA v2 Elo rose 260 point The same report frames Muse Spark 1.2 as among the most cost-efficient models at its intelligence level, at $0.40 per Intelligence Index task at Meta's unchanged $1.25/$4.25 per 1M token pricing — cheaper than GPT-5.6 Terra, Kimi K3, and GPT-5.5, with cost per task up from $0.29 on 1.1 due to a 53% increase in input to