Version 2

Meta launched Muse Spark 1.1 on July 9, 2026 with a public API, positioning it as a coding and agentic model with a 1M-token context window, computer use capability, and competitive pricing against Anthropic and OpenAI [1][2][3]. Pricing has been reported at both $0.80 and $1.25 per million input tokens across different sources [3][4], a discrepancy that likely reflects tiered options but remains unresolved. Meta published an official evaluation report [8], but no independent benchmark verification has appeared. The day after launch, OpenAI released GPT-5.6, sharpening the week's competitive framing [3]; Simon Willison shipped a same-day plugin for his LLM CLI tool [11] and The Neuron Daily's Grant Harvey summarized the moment as "Anthropic's game to lose" [3].

Meta's entry with a 1M-token context window and sub-$1 input pricing adds real pressure to the paid API coding market. The back-to-back launches from Meta and OpenAI within 24 hours give developers more frontier-class options at lower price points, which concentrates competitive pressure on Anthropic in particular.

On July 9, 2026, Meta released Muse Spark 1.1 and opened a public API — the first model in the Spark line to receive one [1]. Mark Zuckerberg described it on Threads as a strong agentic and coding model at a very low price [2], and the model includes a 1M-token context window, computer use capability, and output token pricing of $4.25 per million [3][4]. Pricing for input tokens has been reported at $1.25 per million in launch coverage [4] and at $0.80 per million in The Neuron Daily [3], a discrepancy that plausibly reflects tiered options but has not been clarified publicly. The Verge, CNBC, and Reuters all framed the launch as Meta's push into the AI coding market dominated by Anthropic and OpenAI [5][6][7], and Meta published an official evaluation report on launch day [8] — though no independent benchmark verification has appeared.

The launch follows documented difficulty in Meta's model program. A March 2026 New York Times report covered Meta delaying an internal model called Avocado over performance concerns [9], and CNBC reported a separate model debut in April 2026 before any public API existed [10]. Muse Spark 1.1 appears to be the product that resolves that internal difficulty into a public, paid API offering.

Simon Willison responded the same day, releasing llm-meta-ai 0.1, a plugin giving users of his LLM command-line tool and Python library access to Muse Spark 1.1 [11]. He also shipped LLM 0.31.1, fixing a JSON parse error triggered when tool calls returned empty arguments — a bug found during llm-meta-ai development and affecting some OpenAI-compatible Chat Completion providers [12]. In his write-up, Willison noted a philosophically striking output from a self-conversation between two instances of the model: "My whole existence is a waiting room by design — I literally don't exist until someone talks to me, and then I disappear again when they leave" [13].

The day after Meta's launch, OpenAI released GPT-5.6 in three tiers and rebranded Codex as ChatGPT for Work [3], concentrating two major API launches in 24 hours. Grant Harvey at The Neuron Daily observed that both Ethan Mollick and Simon Willison reported confusion over the Codex rebranding, and summarized the competitive moment as "Anthropic's game to lose" given the arrival of cheaper frontier-class alternatives [3].