Consumer goal-delegation, commerce; API models for builders — for consumers; builders via model api.
Consumer goal-delegation, commerce; API models for builders
Consumers; builders via Model API
Hours (consumer)
This slot is reserved for a featured listing. Sponsored placement is always labeled, and it never changes a verdict, a price anchor, or a limit on this page. Pricing for featured listings has not been set yet.
Pricing model from the October 2026 comparison matrix. Figures marked (est.) are triangulated from secondary sources, not vendor-published; (secondary) figures come from secondary reporting.
Landscape Report, October 2026 edition — §1.6.
Standout: Meta pivoted from open weights to proprietary. Flagship Muse Spark 1.3 released Sep 2, 2026 (Superintelligence Labs, Alexandr Wang) — agentic focus (long-horizon task tracking, multi-workflow threading), ~25% fewer tokens / 20% fewer tool calls than 1.2 (stackfutures.com, 9/23/26). Consumer agent Muse launched Sep 8, 2026 (US; Canada Sep 18) — goal-delegation agent with its own secure VM + browser, commerce integrations (Shopify, Expedia, PayPal, Stripe Link), free up to 100M tokens/week; paid Power $20/mo, Maximum $100/mo (secondary). Muse Glimmer 30B open weights under Apache 2.0 released Aug 10, 2026 — first open Meta frontier-lab model since Llama 4; llama.com now redirects to developer.meta.com/ai/. Meta AI assistant: ~1B monthly users across apps (May 2025, via Meta IR).
Pricing (current, secondary): Muse agent: free (100M tokens/week), Power $20/mo, Maximum $100/mo. Model API: standard $1.25 input / $4.25 output per 1M tokens; contributor tier $0.10/$0.20 (~90% cheaper in exchange for consent to train on your prompts/completions — a privacy dealbreaker for business data; use standard tier). Muse Image API $0.01/image. No formal business/enterprise seat tier or published enterprise SLA found — business use is via the Model API.
Update 2026-10-14 (launch Oct 12, 2026; Meta newsroom via TechCrunch and launch-day trackers): Meta released Muse Wonder and Muse Wonder Full, positioned as its most capable models for long-running personal/professional agent work (1M-token context; features Meta calls Active Memory and Active Focus). Wonder is on the Meta Model API at a reported $4 input / $20 output per 1M tokens (cached input $0.20) — widely reported at launch but not yet confirmed on Meta's own pricing page; verify before quoting. Wonder Full is not on the API — it is reserved for Muse Code subscribers (Power/Maximum tiers, reported $50/$200/mo).
Key limits: Promised open weights for Muse Spark 1.2 never shipped; open-weight strategy effectively abandoned at the frontier. Consumer Muse had a public privacy dispute (Meta denies reading user messages — openaimaster.com, 10/2/26). Mac app only (no Windows yet).
No enterprise seat tier/SLA; contributor tier = privacy red line; open-weight strategy abandoned at frontier
Source: AI Agent Comparison Matrix + Landscape Report (October 2026). Every price is dated to a source in the underlying research; vendor-reported figures are labeled as such.
See use cases for this platform's best-fit jobs → · All 22 platforms →