World Wire

Perplexity has open-sourced SPACE , the sandbox platform behind every task in Perplexity Computer. The Perplexity SPACE sandbox gives each agent task its own Firecracker microVM. Credentials and outbound traffic are controlled from outside that VM. The code ships under Apache 2.0 on GitHub . For teams building coding or browsing agents, this is a production-tested isolation layer they can self-host. TL;DR Size: Rust codebase, release 0.8.0 , the public baseline. HTTP API plus server-sent events. Runs on: A Linux x86_64 host with KVM , /dev/kvm , btrfs and Docker. Local Kind (Kubernetes) or direct mode. No macOS. Performance: In production, median create latency fell from 185 ms to 60 ms versus Perplexity’s previous provider. Best: 0 of 108 VM-to-host escape attempts succeeded across 9 frontier models. Worst: Before fixes, agents bypassed the egress policy in 11 of 54 partial-network runs. What is Perplexity SPACE? SPACE (Sandboxed Platfor...

Liquid AI has released d1-ultra , its most capable decision model to date. The Liquid AI d1-ultra decision model reads text, images, or both. It answers typed questions with a probability for each possible answer. It does this in one forward pass and generates no output tokens. For developers, that means classification, routing and scoring calls without paying for generated text. It is available now on the Liquid API at $0.042 per million input tokens. TL;DR Size: Parameters and context length not disclosed. Built on a larger open-weight backbone than d1. Runs on: Hosted Liquid API only. No downloadable weights or quantized variants announced. Performance: Leads all 8 categories in Liquid’s 112-evaluation suite, per Liquid. Best: 65 on the Artificial Analysis Decision Index, ranked #1 (Liquid-reported). Worst: No public per-benchmark breakdown or parameter count yet; claims are vendor-reported. What is a decision model? A decision model ans...

Meta Superintelligence Labs has released Muse Spark Open , the open-weight version of its Muse Spark model. Muse Spark Open is an 816B-parameter Mixture-of-Experts model that activates 42B parameters per token. It takes text, image, video and audio input, returns text, and holds a 1M-token context window. Meta team claims it to be the most capable US open-weight model to date. For developers, the real news is deployment: the NVFP4 build runs on 2 Apple M5 Ultra Mac Studios. TL;DR Size: 816B total / 42B active parameters (256 experts, 8 active per token); 1M-token context. Runs on: 2 Mac Studios (M5 Ultra, 512 GB each) via MLX; 4 Blackwell Ultra GPUs or 8 H200s with NVFP4; about 300 GB setups via Unsloth 2-bit GGUF. Performance: Strong on agentic tool use; mid-pack among large open models on coding. Best: MCP-Atlas 89.2, the top score in Meta’s comparison table. Worst: Terminal-Bench 2.1 76.4, behind Kimi K3 (85.0), GLM 5.3 (83.9) and Qwen3.8 (82.0)....

TypeSafe AI has released Jev 1.14 , a new version of its System One decision model. It is live now under the jev-preview alias. It adds 3 experimental ‘danger zone’ features: images, a box type and longer context. For developers, Jev 1.14 pushes a typed decision model toward vision tasks like inspection and localization. TypeSafe shares a note that none of the 3 features are ready for production yet. TL;DR: Jev 1.14 Size: Parameters not disclosed. Context is 64k tokens per request and 32k per question by default, or 262k with an opt-in header. Runs on: TypeSafe’s hosted API only ( POST /v1/systemone ). No open weights. Hardware and quantized variants not disclosed. Performance: Small gains everywhere, largest in multilingual support (TypeSafe’s claim, no public scores). Best: In TypeSafe’s Box test, 4 edge-check questions caught 30 of 36 cut-off crops. Worst: Images and Box are not production-ready, and Box has no “not...