What happened
On Thursday Meta publicly launched Muse Spark 1.1, a multimodal model aimed at agentic coding. The first Spark version was announced in April, so this is the follow-up, and it is the one people can actually use.
Meta says the model can reason across multiple steps, work through complex processes, manage digital workflows and push new features into enterprise systems. In plain terms: it is meant to take a task and keep going without a human holding its hand.
The pitch centers on big agentic workloads, bug fixing and large code migrations. That is the unglamorous, expensive work that companies are increasingly happy to hand to a machine.
Then there is the price. According to Reuters, Meta charges $1.25 for each million tokens sent in and $4.25 for each million generated, so developers can compare the bill directly.
That lands close to Anthropic's Claude Haiku 4.5 and OpenAI's GPT-5.6 Luna, though slightly above both.
Why it matters
Meta is late here. Anthropic and OpenAI have offered comparable coding models for quite some time, and the field is crowded.
Late does not mean harmless. Cost of usage remains a constant battleground in AI, and Meta is pricing in the same neighborhood as the incumbents' lighter models. For a company running thousands of coding tasks, small gaps per token add up fast.
Meta is not first, so it is competing on price and on how much agentic work the model can take.
Slightly above the rivals' prices is also a notable choice. Meta is not trying to undercut everyone into the floor. It is saying: same ballpark, come try ours.
Whether Spark 1.1 actually matches the rivals on quality is a different question, and a price list does not answer it.
What to do with it
- Run your own bake-off.Take a real bug backlog or a small migration and feed it to Spark 1.1 and your current tool; compare results, not slides.
- Do the token math.At $1.25 in and $4.25 out per million tokens, estimate your monthly volume before you celebrate any discount.
- Keep your stack swappable.If a third serious player keeps the price pressure on, you want to switch models without rewriting your pipeline.
- Read the vendor claims as claims.Meta says it handles workflows and enterprise deployments; check that against your own systems first.
Ciao, sapiens. The robots now have three competing employers, and you still have no bargaining power in that sentence.


