Meta launches Muse Ember, a low-cost agent model benchmarking Haiku

Meta has released Muse Ember, a lightweight model aimed at high-volume agent work such as tool calls, coding subtasks, routing and long-document parsing. Meta says it matches Claude Haiku 5.5 on the Artificial Analysis intelligence index while costing roughly one-third per task, and it is available now through Meta Model API and Muse Code.

Meta AI has released Muse Ember, which it describes as its smallest and most efficient model, built for the work agents repeat all day: tool calls, coding subtasks, routing, extraction and long document parsing. According to Simon Willison, pricing on shorter prompts matches GPT-6 Luna and Claude Haiku 5.5 at $0.10 per million input tokens.

Meta has released Muse Ember, its smallest model, with a 1M-token context window aimed at tool calls, coding subtasks and document parsing. According to MarkTechPost, it matches Claude Haiku 5.5 on the Artificial Analysis Intelligence Index at roughly one-third the cost per task.

Meta has released Muse Ember, which it calls its smallest and most efficient model yet, with a 1 million token context window and native understanding of text, images, audio and video. According to Artificial Analysis, it scores 35.8 at its highest reasoning setting, close to Claude Haiku 5.5, while Meta positions it for low-cost, high-token agent workflows.

Meta has released Muse Ember, a small, efficiency-focused model now available in Muse Code and through the Meta Model API under the ID muse-ember-1.0, aimed at repetitive agent work such as browsing and terminal operations. Benchmarks posted on Hacker News put it first on six of seven agent tests and tied with GPT-6 Luna on the seventh, though neither Meta nor independent testers have confirmed those numbers.

Meta has introduced Muse Ember, an entry-level AI model that it says costs significantly less than competing offerings and is aimed at powering cost-efficient AI agents, SiliconANGLE reported. Meta said it is the first model trained using Muse Wonder, its flagship large language model, but did not specify which part of the workflow Muse Wonder automated.

Meta has released Muse Ember, its smallest and most efficient model yet, aimed at complex reasoning and repetitive agent tasks with a 1M-token context window and text, image, audio and video input. Meta says it matches Claude Haiku 5.5 on the Artificial Analysis leaderboard at one-third the cost per task, and it is available now through Meta Model API, Google Cloud and Oracle Cloud at $0.10 per 1M input tokens.

Meta released Muse Ember on Nov 9 local time, positioning it as its smallest and most efficient model for high-frequency agent work such as tool calls and request routing. Third-party evaluator Artificial Analysis scored it 36 on its intelligence index, level with Claude Haiku 5.5 at its top reasoning setting, at about $0.14 per task versus roughly $0.41. It is live via the Meta Model API at $0.10 per million input tokens and $0.50 per million output tokens.

Meta has released Muse Ember, a model that combines multimodal reasoning with a 1M-token context window and low API pricing, aimed at repetitive coding and agent workloads, according to TestingCatalog.

Meta has introduced Muse Ember, its smallest and most efficient model yet, aimed at repetitive agent work such as tool calls, routing and data extraction. Meta claims it delivers more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna, though it falls short on actual programming tasks.

Meta released Muse Ember, its smallest and cheapest model yet, built for repetitive agent jobs such as tool calls, coding subtasks, routing, data extraction and parsing long documents. Input pricing is $0.10 per million tokens, and Meta says it matches Claude Haiku at a third of the cost per task.

Meta Superintelligence Lab released Muse Ember on Monday, calling it the lab's lightest and most cost-efficient AI model yet. Meta says it delivers more intelligence per dollar than Haiku 5.5 and GPT-6 Luna, though those comparisons come from the company and no specific benchmark scores were included.

Artificial Analysis reports that Muse Ember scores 21.1% on AA-Omniscience, the lowest among the similarly intelligent models it compared, largely because it declines to answer 64% of questions. That abstention keeps its hallucination rate at 17.4%.

Artificial Analysis says Muse Ember scores 46 on its Coding Agent Index inside Muse Code, ahead of Claude Haiku 5.5 in Claude Code (37) and GPT-6 Luna in Codex (31). It attributes the lead mainly to SWE-Atlas-QnA-AA and Terminal-Bench 4.0.

Artificial Analysis reports that Muse Ember scores 68.1% on AutomationBench-AA, level with Muse Wonder and GPT-6 Astra. Its coding results are mixed: it leads Claude Haiku 5.5 on Terminal-Bench 4.0 (40.9% vs 32.8%) but trails on FrontierCode 2.0 (27.2% vs 34.9%).

Artificial Analysis notes that Muse Ember (max) matches Claude Haiku 5.5 (max) and GPT-6 Luna (max) on input and output token prices, but its cached input is 60% cheaper. Since cached input dominates cost in its agentic evaluations, Muse Ember ends up cheaper per task than similarly priced models.

Artificial Analysis reports that Muse Ember, Meta's newest efficient-tier model, scores 36 on its Intelligence Index, matching Claude Haiku 5.5 (max) at roughly a third of the cost per task. Input and output token prices are the same as Claude Haiku 5.5.

TestingCatalog reports that Meta's new Muse Ember model is now available on the Meta Model API and in Muse Code. The model is positioned as small and low-cost, with performance said to be on par with Claude 5.5 Haiku.

Meta Superintelligence Lab released Muse Ember, its lightest and cheapest model yet, at 10 cents per million input tokens, aimed at the work agents repeat all day: tool calls, coding subtasks, routing, extraction and long document parsing. Meta says Ember is cheaper and smarter than Anthropic's Haiku and OpenAI's Luna on Artificial Analysis's composite benchmark, and positions it alongside a larger model like Muse Wonder that plans while Ember executes.