MarkTechPost
- New LoRA Adapter Lifts Qwen3.8 Coding Success From 39% to 65%MarkTechPost ·
Hugging Face released Qwen3.8-27B-pi-adapted, an Apache-2.0 LoRA adapter trained with reinforcement learning for the pi coding agent.
- SID AI Releases SID-2 Agentic Search Model With 0.720 NDCG Swarm ModeMarkTechPost ·
SID AI released SID-2, an agentic search model trained with reinforcement learning.
- Anthropic Adds Code Mode to Claude Code: Claude Now Chains Tool Calls in JavaScriptMarkTechPost ·
Anthropic has added Code Mode to Claude Code v2.1.323. The experimental feature lets Claude write JavaScript that chains Bash, Read and Edit calls in 1 step, and keep talking while long commands run.
- Perplexity Computer Adds Multiagent: Up to 16 AI Agents That Cross-Check WorkMarkTechPost ·
Perplexity added Multiagent to Computer, an effort level above Ultra where a lead agent runs up to 16 agents on Claude, GPT and Gemini models that share findings and check each other’s work.
- OpenAI Releases GPT-6.2 Astra and Sol With a 20% Cheaper Astra APIMarkTechPost ·
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026. Astra's API price drops 20% to $8/$40 per million tokens, and it ranks second on the Artificial Analysis Intelligence Index.
- Clearing Up the Confusion on AI Data Retention: Two Standard Designs ExplainedMarkTechPost ·
Zero data retention applies to specific AI endpoints, not whole vendors. This guide explains the 2 standard API designs for data retention, who manages deletion in each, and 6 checks for developers.
- NVIDIA Open-Sources NVKE and NVKR for Managed Kubernetes on GPU InfrastructureMarkTechPost ·
NVIDIA has open-sourced NVIDIA Kubernetes Engine (NVKE) and NVIDIA Kubernetes Recipes (NVKR) under Apache 2.0.
- Sarvam Launches Sarvam Token Factory: Bringing India-Hosted Inference to 8 Open ModelsMarkTechPost ·
Sarvam Token Factory is now open to everyone. It serves Sarvam 105B, GLM 5.3, DeepSeek V4.1 Flash, Kimi K3.1 Pro and Gemma 4 31B from GPU clusters in India through one OpenAI-compatible API.
- Liquid AI Releases d1.2 and d1.2-ultra: Omnimodal Decision Models That Top the AA Decision IndexMarkTechPost ·
Liquid AI released d1.2 and d1.2-ultra, decision models that read text, images and audio and return calibrated probabilities with zero output tokens.
- Meet Meta’s JiTTests: LLM-Generated Tests Designed to Fail Before Risky Code LandsMarkTechPost ·
Meta has moved Catching JiTTests into production.
- Alibaba Releases Qwen4-Max: 2.4T MoE, 71B Active, 1M ContextMarkTechPost ·
Alibaba released Qwen4-Max, a proprietary 2.4T-parameter MoE model with 71B active parameters, native vision and a 1M-token context. It costs $1.5/M input tokens and cuts Work360 cost per task by 62%.
- Hark Releases Handoff², a 2x Faster Computer-Use Agent for Hark ProMarkTechPost ·
Hark has released Handoff², the computer-use agent inside Hark Pro. It averages 2x faster than v1, completes 82% of 478 internal web tasks, and learned to batch actions through latency-penalized RL.
- Bland Launches Tofu-1, a Full-Duplex Speech-to-Speech Model for CallsMarkTechPost ·
Bland has released Tofu-1, a full-duplex speech-to-speech model for noisy, high-stakes phone calls. It is in enterprise research preview, with no published benchmarks, size or pricing yet.
- Cmpnd Releases dspy-wsp: Weight-Space Programming Comes to DSPyMarkTechPost ·
Cmpnd released dspy-wsp, a prototype DSPy optimizer for weight-space programming.
- Stanford Research Open-Sources Memorilla, a Latent Memory Module for Frozen LLMsMarkTechPost ·
Stanford researchers released Memorilla, an open-source memory system that compresses a document collection into 16 memory tokens for a frozen Qwen3-8B.
- Vercel Releases UI SDK: 50+ UI Behaviors for AI Coding AgentsMarkTechPost ·
Vercel Labs released Vercel UI SDK, an Apache-2.0 beta library of 50+ UI behaviors.
- Best AI Video Generation Models in 2026: Elo, Price per Second, Open WeightsMarkTechPost ·
The AI video leaderboard looks very different from 12 months ago. Native audio is now table stakes. Clips run 15–30 seconds. And an open-weights model sits in the top 3.
- Perplexity Releases Decider v1.3: Open-Weight Decision Models Add LocateMarkTechPost ·
Perplexity Decider v1.3 and Flash v1.3 are open-weight decision models that return probabilities, not text.
- Google Releases TPU Sync: Open-Source KV Cache Transfer for TPUsMarkTechPost ·
Google released TPU Sync, an open-source C++ library that moves KV caches and weights across TPU HBM, host DRAM, storage and network.
- Google Releases Gemini 4 Hydrogen: 2.5x Faster Than Gemini 3.8 FlashMarkTechPost ·
Google released Gemini 4 Hydrogen, a fast agentic model with a 1M-token window, 4 thinking levels and $1.50/$7.50 pricing. Google reports a 51 Artificial Analysis score and 57% on Terminal-Bench 4.1.
- Liquid AI Releases LiquidCodec, a 74M-Parameter Streaming Audio CodecMarkTechPost ·
Liquid AI released LiquidCodec, a ~74M-parameter streaming audio codec that powers LFM3-Audio-2B and LFM3-Omni-2B.
- CopilotKit Releases OpenBot 1.0: An MIT-Licensed, Self-Hostable AI Coworker With Built-In GovernanceMarkTechPost ·
CopilotKit released OpenBot 1.0, an MIT-licensed, self-hostable AI coworker workspace.
- AWS Releases AWS Emulator: A Local Copy of 200+ AWS Services Built by AI AgentsMarkTechPost ·
AWS released AWS Emulator, a free, CC0-licensed local emulator of AWS APIs built from 438 API models. Autonomous agents wrote all of its code.
- IBM Announces IBM Quantum Nighthawk r3: A New Processor to Support 10k Quantum Operations on 520 Programmable QubitsMarkTechPost ·
IBM announced Quantum Nighthawk r3, a 520-qubit processor designed for 10,000 quantum operations.
- Google Launches Gemini Decisions API With Gemini Decision 1 at $0.10/M TokensMarkTechPost ·
Google launched the Gemini Decisions API, powered by Gemini Decision 1.
- Google DeepMind Releases AlphaGenome 2 for Cell-Type-Specific Variant EffectsMarkTechPost ·
Google DeepMind released AlphaGenome 2, a conditional DNA sequence model that predicts regulatory activity and variant effects per cell type.
- Marktechpost AI Launches GPU Leaderboard and the Deploy It Chrome ExtensionMarkTechPost ·
Marktechpost Launches GPU Leaderboard and the Deploy It Chrome Extension
- MBZUAI’s IFM Releases PAN 2, an Open 29B World Model for Physical AIMarkTechPost ·
The Institute of Foundation Models at MBZUAI released PAN 2, an Apache 2.0 world model that couples an LLM planner with a 29B diffusion transformer to simulate up to 20 seconds of action outcomes as video.
- Fish Audio Releases Live 3, a Full-Duplex Voice Model That Reacts in 186 msMarkTechPost ·
Fish Audio released Live 3, a full-duplex voice model for voice agents. It listens while it speaks, separates “mm-hmm” from real interruptions, and can react in as little as 186 ms.
- Meta FAIR Releases e-means: An Algorithm Beating k-means in 10x Fewer EpochsMarkTechPost ·
Meta FAIR’s e-means is a mini-batch k-means replacement shipped in FAISS 1.16.0. It reaches Lloyd’s 25-epoch loss in 1.4 to 2.6 epochs, with 10-45x GPU speedups and no extra hyperparameters.
- Zyphra Releases ZONOS3-Mini: Open-Source TTS With 100M Active ParametersMarkTechPost ·
Zyphra released ZONOS3-Mini, an Apache 2.0 text-to-speech model with 800M total and 100M active parameters. It clones voices, steers delivery word by word and runs in real time in the browser
- Anthropic Launches Claude Fable 5.2 and Mythos 5.2: Same Model, Different SafeguardsMarkTechPost ·
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, one model with two safeguard levels.
- Meta Releases Muse Ember, a 1M-Context Model for High-Volume Agent WorkMarkTechPost ·
Meta Muse Ember is Meta’s smallest model, built for tool calls, coding subtasks and document parsing.
- Google Unveils TPU 8i Max and TPU 8tx: A Dual-Chip TPU and a Confidential TPUMarkTechPost ·
Google added TPU 8i Max and TPU 8tx to its eighth-generation TPU lineup. TPU 8i Max packs 2 TPU 8i chips per package for long-context agents. TPU 8tx adds hardware confidential inference.
- Baidu’s ERNIE Evolve Automated Research System Beats NanoChat and NanoGPT RecordsMarkTechPost ·
Baidu’s ERNIE Evolve is an automated research system that edits code, runs experiments and keeps only verified gains.
- Reason Machines Open-Sources Aspen, a Git Server Built for AI AgentsMarkTechPost ·
Reason Machines has open-sourced Aspen, a self-hosted Git server for AI agents. It follows OpenAI’s Pando design and sustained 567 pushes per second on one 8-core server in the company’s own test.
- 0sec Labs Open-Sources 0, an AI Security Harness With Web UI, CLI and MCPMarkTechPost ·
0sec Labs has open-sourced 0, an AI security harness for code review, pentesting, reverse engineering and fixes.
- Google AI Releases Elephant-Goldfish Model Plugin for Spec-First CodingMarkTechPost ·
Google's Elephant-Goldfish Model plugin makes coding agents draft specs, then runs 5 zero-context review stages before any code.
- Which Routing Does Each MoE Model Use? DeepSeek, Qwen, Mixtral, Llama 4, gpt-oss ComparedMarkTechPost ·
A config-verified comparison of how 10 open MoE models route tokens: total and active parameters, expert counts, top-k, shared experts, gating functions and load-balancing methods.
- MoE Routing Strategies Explained: Top-K, Expert-Choice, Sigmoid Gating and Dynamic Routing (2026 Guide)MarkTechPost ·
MoE routing strategies decide which experts process each token. This guide explains 7 methods, from Mixtral's top-2 softmax to DeepSeek-V4's hash routing, with the models that use each one.
- Nous Research Releases Arete-27B that Cuts Qwen3.8-27B Propaganda From 22% to 0.9%MarkTechPost ·
Nous Research released Arete-27B, an Apache 2.0 fine-tune of Qwen3.8-27B. On Nous’s 803-prompt eval, propaganda answers fell from 22% to 0.9% while Hermes Bench stayed flat at 65.60%.
- How to Run NVIDIA Dory 4B with llama.cpp: Loop by Loop GuideMarkTechPost ·
Explore a comprehensive coding guide for NVIDIA's Dory 1.0 4B looped language model.
- Buoyant Releases Kinference: An Open Source Inference Control Plane for Kubernetes and vLLMMarkTechPost ·
Buoyant has released Kinference, an Apache 2.0 inference control plane for vLLM on Kubernetes.
- Zyphra ZYFA: FP8 FlashAttention for AMD MI355X, Up to 1.76x FasterMarkTechPost ·
Zyphra released ZYFA, open-source FP8 and BF16 prefill attention kernels for AMD Instinct MI355X.
- Thinking Machines Releases Inkling 2 Small, a 594B MoE With 17B ActiveMarkTechPost ·
Thinking Machines Lab released Inkling 2 Small, a 594B-parameter MoE with 17B active parameters.
- Google Releases Gemini Omni 1.2 Flash: 15-Second AI Video at $0.07 per SecondMarkTechPost ·
Google released Gemini Omni 1.2 Flash, a video generation and editing model with native 15-second clips, multi-turn conversational editing and a 30% price cut to $0.07 per second of 720p video.
- Baidu Releases ERNIE 5.5 Flash Base: A 280B MoE Base Model With 14B Active ParametersMarkTechPost ·
Baidu has released ERNIE 5.5 Flash Base, a 280B sparse MoE base model with 14B active parameters, trained on 33.5T tokens and published before any mid- or post-training.
- Together AI Releases Together Compiler: Agents Make Inference Up to 2.6x FasterMarkTechPost ·
Together AI released Together Compiler, an agent-driven loop that keeps re-optimizing the inference forward pass as traffic shifts.
- OpenAI Launches GPT-Image-2.6 Flare and Sunburst: Adds GIFs, SVG Export and 6-Second API GenerationMarkTechPost ·
OpenAI released Images 2.6 and the GPT-Image-2.6 Flare and Sunburst API models.
- Google AI Research Releases Agile Deliberation Framework for Gray-Area Image ClassifiersMarkTechPost ·
Google Research released Agile Deliberation, a human-in-the-loop framework that helps users define subjective visual concepts through borderline images.