News · Leaderboard
Read what just changed, and see who is ahead. Those are the two doors on this page.
objekts Production Desk is the first-party production preflight interface for commercial AI/VFX and branded visual work. It checks reference authority, hard locks, asset fitness, continuity and feasibility; produces directional estimates and briefs; and can hand an approved brief to the human objekts production team.
On [2026-10-03](https://huggingface.co/blog/microsoft/thinkingbox), Microsoft Copilot Studio and Hugging Face released **ThinkingBox** plus **ThinkingBox-Bench** (507 tasks × 20 trials): isolated MCP sessions graded by **executable terminal-state/side-effect checks**, wired into [OpenEnv](https://huggingface.co/docs/openenv/environments/thinkingbox). Vendor board: Claude Opus 5.5 **67.16%** pass@1; Kimi-K3 **93.89%** pass@20 but only **13.41%** all-20 — one success ≠ reliability. MIT framework [microsoft/thinkingbox](https://github.com/microsoft/thinkingbox); CDLA data [microsoft/thinkingbox-data](https://github.com/microsoft/thinkingbox-data); paper [arXiv:2608.19741](https://arxiv.org/abs/2608.19741).
Anthropic blog [Code Review](https://claude.com/blog/code-review) (Last-Modified **2026-10-02**) and [docs](https://code.claude.com/docs/en/code-review): multi-agent **Code Review** research preview for Claude Code **Team/Enterprise**. Parallel find→verify→rank; overview + inline comments; never approves/blocks. Vendor stats: substantive comments **16%→54%**; large PRs **84%** with findings; ~**$15–25**/review, ~**20 min**. Local `/code-review` remains for other plans.
GitHub [dsh-v0.2.1-alpha.1](https://github.com/deepseek-ai/deepseek-harness/releases/tag/dsh-v0.2.1-alpha.1) (**2026-10-03T06:42:19Z**) adds an **experimental Claude Code Mods compatibility layer** to MIT **DeepSeek Harness** (~**243k** stars)—explicitly to test that Mods APIs are broadly a subset of dsh plugins, **not full compatibility**—continuing v0.2 desktop apps. Get it at [deepseek.com/harness](https://deepseek.com/harness) or `npx @deepseek-ai/dsh web`.
arXiv [2610.02163](https://arxiv.org/abs/2610.02163) (PDF stamp **2026-10-02**) and [autocompact.github.io](https://autocompact.github.io/) introduce **AutoCompact**: a proactive `compact()` action in a terminal REPL scaffold, trained via judge-corrected on-policy SFT then outcome GRPO so the policy learns when to compact, what to keep, and how to continue. Base **Qwen3-Coder-30B-A3B-Instruct**; SWE-bench Verified **39.6%** vs **30.4%** base (+**9.2** pp); SWE-PolyBench Verified **24.5%** vs **19.5%** (+**5.0** pp). **No public weights/code repo found.**
On **2 Oct 2026** Apple posted on [Developer News](https://developer.apple.com/news/?id=p6zjojqw) that macOS **Full Disk Access** will get additional controls so grants require **very explicit user action**. The post names **AI agents** as the reason risk will grow substantially. No OS version or ship date was announced.
On **2 Oct 2026** NVIDIA blogged [DGX Spark 64GB](https://blogs.nvidia.com/blog/local-ai-dgx-spark-64gb-sync/): on sale **Friday 23 Oct** via Acer/ASUS/Dell/Gigabyte/HP/MSI from **$4,999**. Same **GB10 Grace Blackwell**, DGX OS, and full NVIDIA AI stack; **64GB** unified memory for local agents up to ~**100B** params. Two units cluster over ConnectX-7+QSFP to **128GB** (~**200B**); Qwen 3.8 27B test showed up to ~**1.7×** vs one node.
On **2 Oct 2026** OpenRouter shipped [Model Router Benchmarks](https://openrouter.ai/benchmarks/routers): seven model routers (Auto, Jev, NVIDIA Switchyard, Unbiased Pareto, Sakana Fugu, and variants) compared on GPQA Diamond, Tau3 Banking, MMMU Pro Vision, SWE Atlas QA, Deep SWE, and Terminal Bench 2.1. The Router Index defaults to quality 60% / time 20% / cost 20%. `openrouter/pareto-code` and Fusion are not on this board.
On **3 Oct 2026**, Aleph Alpha released open-weight **Kolibri-1** ([HF](https://huggingface.co/Aleph-Alpha/Kolibri-1), **Apache-2.0**): 78B MoE / ~3.46B active, bilingual DE/EN, 262k native context (1M extrapolated), reasoning effort + tools. Official card: LiveCodeBench v6 **85.9**, SWE-Bench Verified **66.4**, TerminalBench 2.1 **27.7**. Serve via `aleph-alpha-inference` + vLLM.
vllm serve Aleph-Alpha/Kolibri-1 with kolibri1 parsersOpen-source [Louis-CFM/coucou](https://github.com/Louis-CFM/coucou) (MIT, ★3146) parks Claude Code/Codex/Cursor/Gemini CLI/Antigravity sessions in the Mac notch (top-edge island on Windows/Linux), with in-island Allow/Deny/Always approvals, live diffs, file drop, and service pills. macOS **v0.1.3** shipped 2026-10-03; Linux **0.1.1 beta** is out; Windows installer is paused for a Defender false positive.
coucou_agent (AGENTS.md)Large language models are rapidly deployed across cybersecurity workflows to translate analysts' intent into command-line interface (CLI) executions. However, existing benchmarks emphasize either high-level knowledge QA or unconstrained agent rollouts, failing to measure precise parameter binding across real-world security tooling where minor syntax glitches invalidate execution. Researchers from MBZUAI introduce KaliBench, a fine-grained natural-language-to-CLI benchmark on Kali Linux spanning 8,504 query-command pairs across 1,642 tools, 23 capability dimensions, and 5 security phases. Auditing reveals no open-weight model surpasses 42% exact accuracy without tool hints. Furthermore, reinforcement learning with KaliBench's runtime-free verifiable rewards elevates an 8B model to rival a 685B MoE frontier model.
Conventional LLM agents emit reasoning thoughts and actions token-by-token, making extended interactions slow and compute-heavy. While Jev-style probabilistic models deliver rapid predictions over finite action domains, they mandate pre-specified, static fields, precluding autonomous adaptation in open-ended language tasks. Shanghai Jiao Tong University researchers introduce JevSpawn, a compositional policy framework bridging natural language instructions with finite probabilistic exploration. JevSpawn couples parallel action spawning with feedback-driven branch selection, representation revision, and recovery from cached alternatives. By sharing prefix KV caches and action structures, JevSpawn eliminates repeated context generation without model training, systematically outperforming seven leading agent baselines across eight benchmarks.
On-policy self-distillation (OPSD) trains mathematical reasoning models by deploying a privileged teacher that observes ground-truth solutions to supervise student-sampled prefixes. Standard OPSD fixes the teacher parameters across all states, leaving valuable supervision untapped. Researchers from Tencent WeChat and Tencent AI Lab propose Neighborhood OPSD (N-OPSD). They discover that local parameter perturbations in the privileged teacher reveal complementary, reference-aligned corrections across distinct token positions. Offline greedy pruning compiles a compact pool of frozen neighbor experts, while an online MaxPeak and quantile router dynamically selects the optimal expert distribution. Across AIME 2024, AIME 2025, and HMMT 2025, N-OPSD improves Average@12 by up to 2.75 points on Qwen3-1.7B, 4B, and 8B models without any inference overhead.
Official Antigravity Models docs now list Claude Sonnet 5.5 (thinking) and Claude Opus 5.5 (thinking): Free/Plus/Enterprise ❌; Google AI Pro non-trial only ✅; Ultra ✅. Footnote: Claude 4.6 and GPT-OSS-120b remove on 2026-11-02. No separate changelog post—docs table is the source. No invented Antigravity SWE scores.
At DevDay (2026-09-29) OpenAI shipped reusable Codex Cloud environments: publish a project setup (repos, deps, tools, network secrets), then each task runs in its own VM workspace and can continue while your laptop sleeps. Docs: learn.chatgpt.com/docs/cloud and Help Center. Rolling out to Plus/Pro and workspace plans; saved VM state recoverable ~7 days by default. No invented SWE scores.
stablyai/orca positions itself as an Agent Development Environment (ADE): parallel CLI coding agents (Claude Code, Codex, OpenCode, Pi, …) each in an isolated git worktree, with desktop, mobile companion, and remote hosts. Verified now: ★84,136 / 5,424 forks; MIT; desktop v1.4.219 (2026-10-02); Android mobile-android-v0.0.52 (2026-10-03). Site onOrca.dev. No official SWE leaderboard—no invented scores.
Before invoking tools, agentic LLMs navigate a K-way action space: executing calls, clarifying ambiguities, answering directly, or declining. While internal activation steering attempts to align pre-execution tool decisions, aggregate metrics obscure where manipulated latent states land and the severe collateral damage they inflict. Researchers from Aberdeen and Oxford introduce SAKIKO, an auditing framework formalizing representation repair via directional error discovery, router-conditioned intervention, destination-resolved verification, and prospectively frozen statistical licensing. Crucially, destination auditing demonstrates that behavioral movement does not equal repair: an intervention scoring a net +55 gain corrupted more than half of baseline-correct decisions, establishing the necessity of outcome-resolved adjudication.
Expanding text embedding models to multimodal domains typically triggers catastrophic forgetting in text retrieval, with existing omni-modal embedders ballooning past multi-billion parameters to compensate. MBZUAI researchers introduce Omni-Embed-Mini, a compact 0.9B architecture mapping text, speech, audio, image, video, and rich documents into a single shared cosine space. Crucially, text-side weights remain bit-identical to the base model, guaranteeing zero regression on text retrieval (49.57 nDCG@10 on MTEB-v2 BEIR-8). Powered by dense cascaded caption distillation, Matryoshka SigLIP loss, and online hybrid hard-negative mining, Omni-Embed-Mini is 2.7x to 9.5x smaller than competing open models, while its 2.3B variant surpasses proprietary gemini-embedding-2 on cross-modality averages.
Real-world embodied tasks require agents to operate physical tools under geometric constraints while tracking object state transitions. Despite strong perception benchmarks, current multimodal video LLMs struggle with tool-centric physical reasoning. NTU S-Lab and collaborators introduce EgoTools, the first comprehensive suite for egocentric tool-use reasoning. It features EgoTools-Data (100 hours of synchronized egocentric tool videos with 3D annotations, audio, and dense causal narrations) and EgoTools-Bench (1,000 QA pairs across perception, geometry, procedural progress, and causal inference). Probing reveals frontier models struggle with physical grounding (Gemini-3.1-Pro scores only 51.7% on grounding), while supervised fine-tuning on EgoTools-Data lifts Qwen3-VL-8B-Instruct accuracy from 50.0% to 60.9%.
EverMind-AI/Raven orchestrates built-in Research/Code/Design/Oncall plus third-party harnesses (Claude Code, Codex, Hermes, Pi, …) via a Host Agent DAG with admission checks, and frames controlled self-evolution of Raven’s own harness (RSI). Verified: ★5097 / 145 forks; v0.2.3 (2026-09-27); arXiv 2609.33439. Site: Raven-Research 76.5% DeepResearch Mixed; paper: MAOB Exact Match +10.4/+10.5 pp vs strongest baseline. Pre-alpha.