Grok 4.6 출시와 에이전트 성능, Android 수어 입력, 오픈 웨이트 모델 실행 경로

Elon Musk

@elonmusk

Grok 4.6 is now out Smart, fast & amazing bang for buck!

Elon Musk

@elonmusk

Grok 4.6 reaches 1753 ELO

SL2T is our breakthrough sign language-to-text model powering new features for Deaf and hard of hearing users on @Android . Starting with American Sign Language-to-English on Pixel 11, people can sign directly into Gboard and Live Transcribe instead of typing.

SL2T is state-of-the-art on academic benchmarks, plus it's optimized for real-world use like one-handed signing while holding a phone. To protect privacy, it tracks body poses on-device, while our servers translate them into text.

We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer looks like: LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king ): 304B MoE refresh, Terminal-Bench 2.1 jumps 61.8→82.7 over the preview, DeepSWE 7.3→54.4. Closes in on Opus-4.8 on Agents' Last Exam (25.2 vs 25.7). MIT. → Muse-Glimmer-30B, from Meta (they are back!!): their first open agentic model. ~29.6B dense + perception encoder, 131k+ context, built to run fully local, no cloud. Apache 2.0. → Liquid AI LFM2.5-2.6B: 2.69B params, 131k context, 220 tok/s on an M5 Max in under 2.5GB RAM. Competitive with models 4x larger on agentic tasks. → inclusionAI Ling-3.0-flash: 124B total, only 5.1B active, ~12% the size of their old 1T flagship Ring-2.6, matches it on key benchmarks. MIT. → inclusionAI Ling-3.0-tiny: 7.9B total, 1.3B active, 86-90 tok/s on an M4 Pro MacBook at ~8GB peak memory. MIT. → NVIDIA Nemotron-3.5-Lightning-30B-A3B: hybrid Mamba-2+MoE+Attention, up to 1M context, runs on a single H100 or DGX Spark, SWE-bench Verified 52.8. → deepgrove maple-preview: 20B-A1B ternary-weight reasoner, 218 tok/s on a Mac mini M4, 5.3GB checkpoint. MIT. → BigBang-v1 (endless-frontier): fine-tuned from Qwen3.6-35B-A3B via a self-evolving generator/critic synthetic-data loop. Lands aggregate performance between DeepSeek V4 Flash (284B) and V4 Pro (1.6T), at 35B. Apache 2.0. Video → MiniMax-H3: 33B dense omni model, native stereo audio, up to 2K/15s. 3.6k+ likes already. → Minimax-H3-Turbo (lightx2v): Apache-2.0 turbo distillation of H3 for fast inference. → Lightricks LTX-2.5: image-to-video update, custom Gemma-4-12B text encoder, a markedly stronger distilled model. Voice → NVIDIA NemotronLabs VoiceChat-11B: full-duplex speech-to-speech, ~450ms turn-taking, #2 on open VoiceBench, and the first open full-duplex model with live tool-calling mid-conversation. Safety → Mistral Shieldstral-1.0-3B: 3B multimodal guardrail that takes your safety policy as plain text instead of fixed categories. Beats LlamaGuard-4-12B and ShieldGemma-9B on HarmBench (99.4) and ToxicChat (84.1) at a fraction of the size. Apache 2.0.

Congrats to @Alibaba_Qwen on Qwen3.8-2.4T-A95B, one of the largest open-weight models released to date. 2.4T params, 95B active, 512 experts. Day-0 support in vLLM, verified on @NVIDIA and @AMD hardware. A ready-made 4-bit checkpoint per vendor, both out of the box: Inferact/Qwen3.8-2.4T-A95B-NVFP4, 1.32 TiB, one NVIDIA 8xB300 node Inferact/Qwen3.8-2.4T-A95B-MXFP4, 1.45 TiB, one AMD 8xMI355X node No conversion, no calibration on your side. Just vllm serve. Thanks to @Alibaba_Qwen for the weights and the collaboration, @NVIDIAAI and @AIatAMD for the joint kernel engineering, @inferact for the quantized checkpoints and vLLM integration, @digitalocean and @togethercompute for early testing, and the vLLM community.

Expedia recently moved its ranking models to a state-of-the-art Keras 3 setup. Results: 30% faster training, and inference latency decreased by 70%. Read their writeup about the upgrade:

We want AIs to be able to help with work to reduce AI risk. But while models do great in domains where reliable feedback is relatively cheap and abundant, like Math and coding, a lot of work on AI risk isn't like that. Instead, we have to rely on good argumentation to answer questions like "does this experiment tell us anything about future models that are much smarter than humans?" Unfortunately, this kind of work seems much harder to measure (and hence automate). Our team at @redwood_ai developed the Conceptual Reasoning Index (CRI) in collaboration with @AnthropicAI to fix this. Every single data point in the CRI has been manually checked by a researcher on our team to ensure quality. This chart shows the performance of each tested company's highest-scoring model plus Fable 5, Muse Spark 1.2, and Gemini Flash 3.6 which are often their company's frontrunners on other capability benchmarks. A score of 0 corresponds to randomising guessing on all three benchmarks and a score of 100 is the highest possible score on all. We estimate 91 to be the true performance ceiling. More info below. Official leaderboard website which we'll keep up-to-date: https:// conceptualreasoning.ai

Our finance team shared 16 workflows with ChatGPT Work and Codex, from monthly close to forecasting and treasury. Really good read especially if you’re in finance.

witcheer

@witcheer

someone in the community put Hermes Agent on a Raspberry Pi 4 with 2GB of RAM. the Pi 4 is a credit-card-sized computer that costs about $50 and draws a few watts. 2GB is less memory than most phones have. it works because of how Hermes Agent splits the job: the agent runs on the Pi, and the model does not. it talks to models over API, so the heavy lifting happens elsewhere and the Pi only runs the agent loop, memory and tools. this way you get an always-on agent and no VPS costs on a very affordable hardware.