AI Daily · Oct 13, 2026: Meta launches Muse Wonder frontier model family with 1M-token context at $4/$20
Meta launched the Muse Wonder frontier agent family with 1M-token context at $4/$20 per 1M tokens, live on OpenRouter and more; X Subscription bundles Grok, Cursor and X Premium from $8/month.
Showing the latest briefing (Oct 13, 2026). Every day is generated the next morning.
Pick a date
-
Meta launches Muse Wonder frontier model family with 1M-token context
Muse Wonder is a new family of frontier models aimed at personal and professional agentic work, priced at $4 per 1M input tokens and $20 per 1M output tokens ($0.20 cache reads), with a 1M-token context window and text, image, video and PDF input. It is live on OpenRouter, Requesty and Command Code, with day-0 LiteLLM support.
-
X Subscription unifies Grok, Cursor and X Premium from $8/month
The X Subscription bundles Cursor coding agents, Grok chat/create tools, Grok Bot and X Premium perks into one plan starting at $8/month across five tiers, with a shared usage pool; existing Cursor and SuperGrok subscribers can keep their plan or switch in account settings, and paid plans get twice the coding usage on Grok models versus the equivalent Cursor plan.
-
OpenAI's Codex and ChatGPT usage limits reset globally
Tibo Sottiaux of OpenAI said a global reset of Codex and ChatGPT usage limits landed by end of day PST, and teased a simple plan to give everyone significantly more usage, which he hopes to ship soon.
-
Command Code gives DeepSeek V4.1 Flash $100 usage for three days
Command Code announced that DeepSeek V4.1 Flash now comes with $100 of usage for the next three days, with a $10 GOAT plan converting to $100 in credits; after that window the credit amount drops to $80.
-
MiMo-V2.5 models migrate to V2.6 with new pricing
MiMo-V2.5-Pro and MiMo-V2.5 will be routed to MiMo-V2.6-Pro and MiMo-V2.6-Flash starting Oct 14 18:00 SGT and billed at the new models' rates; legacy IDs stop working Oct 21 10:00 SGT.
-
Tencent launches LightVela 24/7 cloud AI assistant with free first month
Tencent introduced LightVela, a 24/7 cloud AI assistant usable via Telegram, WhatsApp and Discord bots with per-bot memory and schedules; the first month is free.
-
ScaleDown constrained-decoding models launch on OpenRouter
OpenRouter now hosts a family of constrained-decoding models from ScaleDown, each built for one task: Summarize condenses long text, images and PDFs into short summaries for context compaction, while Extract takes plain-English field descriptions and returns clean JSON with confidence scores.
-
Google's Embedding Gemma 2 runs free and locally on Mac
Google's Embedding Gemma 2 is described as 100% free and runs locally on Mac hardware.
-
oQe quantized models updated in oMLX 0.7.1 with lower KL divergence
All oQe models were updated with improved oQe quantization, available from oMLX 0.7.1. Compared with previous oQe builds, KL divergence versus the original BF16 model is 11.6–18.1% lower on code and 10.2–13.8% lower on prose.