# Muse Wonder ships without scores, Anthropic tests self-copying > Meta launches Muse Wonder with no published numbers, Anthropic measures small-model self-replication on TV boxes, plus Qwen, MiniCPM and NVIDIA. Published 13 Oct 2026. Coverage: 12 Oct 2026 to 13 Oct 2026. By Ali Asif. ## Items 1. **Muse Wonder** (Meta, Frontier): A frontier model aimed at personal and professional agents, with Active Memory and Active Focus. API only, and no benchmark numbers are published. [Primary source](https://research.meta.ai/blog/muse-wonder) 2. **AI models may be capable of self-replicating without access to a datacenter** (Anthropic, Frontier): Frontier Red Team results on small models copying themselves across cheap IoT hardware, with the setup caveats stated in the post. [Primary source](https://www.anthropic.com/research/self-replicating-ai) 3. **Qwen-Image-2.1-Turbo** (Qwen, Image, video, world): An 8-step checkpoint of the 7B Qwen-Image-2.1 for text-to-image and editing, loadable directly in Diffusers. [Primary source](https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo) 4. **MiniCPM-V 4.7 35B-A3B** (OpenBMB, Open weights): The first mixture-of-experts model in the MiniCPM-V series, with a claimed 13+ tokens per second in 3 GB on an iPhone 17 Pro Max. [Primary source](https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B) 5. **vLLM v0.31.0** (vLLM, Open weights): 717 commits from 307 contributors, with a new preload command that keeps weights resident across engine restarts. [Primary source](https://github.com/vllm-project/vllm/releases/tag/v0.31.0) 6. **ReplaySSM: Rethinking State Storage for Faster Hybrid-Model Decoding** (NVIDIA, Science): Stores state-space layer state as a checkpoint plus recent inputs, cutting memory traffic and simplifying speculative decoding rollback. [Primary source](https://developer.nvidia.com/blog/replayssm-rethinking-state-storage-for-faster-hybrid-model-decoding) 7. **Codex 0.162.0** (OpenAI, Agents and coding): Adds managed Git worktree tools, task pinning in the Command Center and a signed PowerShell installer for Windows. [Primary source](https://github.com/openai/codex/releases/tag/rust-v0.162.0) 8. **Coding-Agent Benchmarks Should Match Their Users' Task Flows** (JetBrains, Agents and coding): Built from 4,782 real IDE agent sessions. Multi-step solving roughly doubled agent cost with no stable change in resolve rate. [Primary source](https://huggingface.co/papers/2610.09633) 9. **Reflex: Visually Grounded Reactive Humanoid Control** (Reflex authors, Robotics): A humanoid that catches thrown boxes from egocentric RGB-D, tested on a real Unitree G1 at 2.5 m and 5 m. [Primary source](https://huggingface.co/papers/2610.12772) Meta launched a frontier model without publishing a score. Anthropic published a pile of scores and then told you how easy the test was. ## Muse Wonder, from Meta Meta announced Muse Wonder on October 12 as a model for personal and professional agents. The pitch rests on two features. Active Memory tracks context across conversations, and Active Focus keeps the model working toward a long goal. You can reach it through the Meta Model API and Muse Code, with a Muse Wonder Full tier for subscribers. There are no open weights and no parameter count. Meta claims state-of-the-art results and cites Harvey's Legal Agent Benchmark, FocusBench and SWE-Marathon, but none of those come with a number. Until someone reproduces one, I read this as a product launch. ## Anthropic on self-replication The Frontier Red Team asked whether small models can copy themselves across cheap IoT devices with no datacenter. A 9B model ran inference on 8 of 9 hardware types. GLM-5.2 reached all 8 nodes in 4 of 5 trials. Gemma4-26B, spread over 15 smart TV boxes at about $55 each, managed 0.39 tokens per second. Setup is slow: 28 to 51 hours for Qwen3.5-9B and 71 to 150 hours for Gemma4-26B-A4B. The number I keep rereading is the one on inherited notes. Success rose from 45% to between 61% and 84% across six generations when each model got notes from the one before. Anthropic projects small-model replication capability by the end of 2026. It also says the test used unauthenticated ADB ports, that a cycle takes days, and that its capability metric is only a proxy. ## Open weights and images Qwen-Image-2.1-Turbo is an 8-step checkpoint of the 7B Qwen-Image-2.1, for text-to-image and editing, and it loads directly in Diffusers. OpenBMB's MiniCPM-V 4.7 35B-A3B is the first mixture-of-experts model in its series. The team claims 13+ tokens per second in 3 GB on an iPhone 17 Pro Max, and a higher overall score than Qwen3.6 35B-A3B in less time. Both are vendor claims. vLLM v0.31.0 landed with 717 commits from 307 contributors, plus a preload command that keeps weights resident across engine restarts. ## Inference and agents NVIDIA and Princeton's Dao AI Lab published ReplaySSM, which stores state-space layer state as a checkpoint plus a buffer of recent inputs. Decoding kernels ran 1.46x to 1.80x faster at batch size 256 on B300 GPUs, and speculative throughput rose 1.85x to 2.10x on 120B-class models. Three of four models stayed within about 2 points of FP32 on GPQA-Diamond. Codex 0.162.0 adds managed Git worktree tools and task pinning. JetBrains studied 4,782 real IDE agent sessions and found only 33% had three or more user messages. In a 700-task SWE-Bench Pro pilot, solving in several steps roughly doubled cost with no stable change in resolve rate. On the robotics side, Reflex catches thrown boxes from egocentric RGB-D. On a real Unitree G1 it succeeds 65% of the time at 2.5 m and 40% at 5 m. ## What I'd watch next Whether Meta publishes any Muse Wonder score on FocusBench or SWE-Marathon. Whether the Anthropic team reruns the test against devices that have real authentication. And whether MiniCPM's phone speed holds up on hardware that isn't the team's own. ## My take Meta launched a frontier model without a single score, and Anthropic published the opposite: hard numbers on small models copying themselves, with every caveat attached. I trust the second post more. ## Sources - https://research.meta.ai/blog/muse-wonder - https://www.anthropic.com/research/self-replicating-ai - https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo - https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B - https://github.com/vllm-project/vllm/releases/tag/v0.31.0 - https://developer.nvidia.com/blog/replayssm-rethinking-state-storage-for-faster-hybrid-model-decoding - https://github.com/openai/codex/releases/tag/rust-v0.162.0 - https://huggingface.co/papers/2610.09633 - https://huggingface.co/papers/2610.12772 Canonical: https://heyaliasif.com/ai/2026-10-13-muse-wonder-self-replication-radar