Meta ships Muse Image API at $0.01 per image
Alexandr Wang is pricing Meta's generation-and-editing model for production volume, seven weeks after its consumer debut.
By Ryan Merket · Published
Primary source: Alexandr Wang on X
Why it matters
Meta is using a flat one-cent price to pull image workloads onto its new model API, turning Muse from an in-app feature into a direct bid for developer adoption.
Alexandr Wang (@alexandr_wang) said in an August 28th post on X that Meta has put Muse Image on the Meta Model API at $0.01 per generated image, extending the model from Meta's consumer products into software developers' production pipelines.
Wang called the offering "one of the best price-to-quality ratios for production volumes." That is Meta's positioning claim. The price is confirmed by the launch announcement, while cross-vendor quality comparisons depend on resolution, editing inputs, quality settings and batch terms.
Wang has served as Meta's chief AI officer since June 2025, according to a Meta regulatory filing. He founded Scale AI in 2016 and ran it until joining Meta. Before Scale, Wang attended MIT and worked on machine-learning infrastructure at Quora, with earlier roles at Hudson River Trading and Addepar, according to his Scale AI biography.
The API release follows Meta's July 7th introduction of Muse Image, when Meta made the model available through the Meta AI app, meta.ai, Instagram Stories in the US and WhatsApp in selected countries. The August release gives developers direct access for applications that need to create, edit or compose images programmatically.
A production price, with caveats
At one cent per output, 100,000 generated images would carry a $1,000 list price. That makes Muse Image viable for jobs where each user, product or advertising variation requires a separate visual, rather than reserving generation for occasional design work.
Meta's rate sits below several current image-generation listings. Google's Gemini API pricing puts Gemini 3.1 Flash Lite Image at $0.0336 for a 1K image under standard processing and $0.0168 in batch. Google's faster Gemini 3.1 Flash Image starts at $0.045 for a 0.5K standard output and rises with resolution.
OpenAI uses token-based pricing for its current GPT-Image-2 API. Its standard rate is $30 per million image-output tokens, with batch processing priced at $15 per million. The resulting charge changes with output dimensions and quality, preventing a clean one-number comparison with Meta's flat advertised price.
Muse Image's actual production economics will also depend on latency, failure rates and how often an application must regenerate an unacceptable result. A cheap image that needs three attempts costs three cents and adds delay. Meta has not supplied an independently audited success rate for real production workloads, so developers will have to measure usable output per dollar against their own prompts and reference assets.
Meta turns Muse into a platform
Muse Image differs from a basic prompt-to-pixel generator in how Meta says it handles complex requests. The model can invoke search and coding tools, refine its own generations and use additional inference-time computation before returning an image. It supports targeted edits and can compose people, objects, styles and settings from multiple reference images.
Meta said at the July launch that Muse Image ranked second in human-preference Arena measurements for text-to-image generation, single-image editing and multi-image editing. Those rankings were a snapshot dated July 5th and were reported by Meta, rather than a standing guarantee of quality as competing models change.
The model can also work with Muse Spark, Meta's reasoning model, allowing the two systems to plan and use tools together. Meta opened the Meta Model API in public preview alongside Muse Spark 1.1 on July 9th. Adding image generation broadens that API from reasoning, coding and computer-use tasks into creative production.
That expansion explains the price. Meta is using inference economics to recruit developers while image generation is becoming a standard feature inside advertising tools, commerce software and consumer applications. Muse Image gives Wang's organization a product that can be measured through API calls and developer adoption, rather than use inside Meta's own apps alone.
For developers, the initial decision is straightforward: whether Muse Image can keep subjects, layouts and text consistent enough that a one-cent generation survives review. The answer will determine whether Meta's price buys distribution or merely cheaper experimentation.