Muse Wonder vs GPT-6: Two Flagships Meet in One Week

Muse Wonder is the model family Meta Superintelligence Labs launched on October 12, comprising the standard Muse Wonder and the more capable Muse Wonder Full. Meta is not releasing the weights: the standard model is sold per token through the Meta Model API, while the stronger version is reserved for paid Muse Code subscribers. Meta highlights two capabilities — Active Memory (retaining conversation history, preferences and habits across sessions) and Active Focus (carrying complex or ambiguous tasks through to completion); the models support 1M-token context and accept text, image, video and audio input.

GPT-6 is the model family OpenAI rolled out across ChatGPT on October 7, alongside Intelligent UI: answers are no longer text-only — ChatGPT automatically combines text, charts, maps, tappable buttons and small tools (such as calculators and bill splitters) based on the question, while still using plain text when it fits best. Plus, Pro, Business and Enterprise tiers use GPT-6 Sol; Free and Go tiers use GPT-6 Luna; Pro's reasoning option stays on GPT-6 Astra without Intelligent UI; the models behind Work and Codex did not change.

The table below places public pricing, independent scores and key features side by side; every figure comes from vendor announcements or third-party evaluations, and each row compares the same dimension.

1. Launch dates: : OpenAI announced GPT-6 on October 7 and began the paid-tier rollout the same day, with Free and Go tiers following from October 8; Meta announced Muse Wonder on October 12.
2. API pricing: : Muse Wonder lists at $4 input / $20 output per million tokens on the Meta Model API, $0.20 for cached input; OpenAI's listed GPT-6-series API pricing is $10 input / $50 output per million tokens for GPT-6 Astra (per Artificial Analysis).
3. Independent index: : on the Artificial Analysis Intelligence Index, Muse Wonder scores 52 and GPT-6 Astra 53; Muse Wonder Full scores 56 but is not on the API — it ships only with Muse Code subscriptions ($50/$200 per month).
4. Vendor-reported performance: : Meta's model card puts Muse Wonder at 78.8 on DeepSWE and 53.1 on SWE-Marathon, above Claude Opus 5.5's 74.2 and 46.3; OpenAI says GPT-6 Instant starts answering web-search questions 44% sooner on average (OpenAI internal evaluation). Both figures are vendor-reported.