Daily AI CatchupMeta Opens Muse Spark 1.1 to Developers at a Quarter of Rival Pricing
Meta released Muse Spark 1.1 via public API preview at $1.25/$4.25 per million input/output tokens — roughly a quarter of what competitors charge — claiming near-state-of-the-art performance on agent reasoning and tool use, beating Opus 4.8 and GPT-5.5 on several agentic benchmarks. It supports splitting jobs across parallel subagents and a 1M-token context window.
Why it matters
💻 Developer · A 1M-token context window with parallel subagent support at a quarter of premium pricing is worth a direct benchmark against your current agentic-workflow model.
📦 Product · If cost has been the blocker on shipping an agentic feature, this pricing tier is worth re-running your unit economics against before deciding it's not viable.
🎨 Design · No direct design impact — this is a backend model release aimed at developers.
📈 Business · Meta undercutting premium pricing by 4x on a benchmark-competitive model puts real pressure on every provider currently charging frontier-tier prices for agentic capability.
🤔 Just Curious · Meta released a new AI model to developers that performs close to the best paid options out there, but costs about a quarter as much to use.
Try this: Benchmark Muse Spark 1.1 against your current agentic-workflow model on a real tool-use task — the pricing gap alone may justify the switch.
Sources: Meta: Introducing Muse Spark API