The latest AI news, in focus.

NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt

NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt

Fix evaluation bias to unlock better models on Amazon Bedrock

Property Finder found that 96.8% of their vision-AI evaluation labels were inherited from the incumbent model, hiding its true accuracy. See how they built unbiased ground truth, applied evolutionary prompt optimization, and selected a more accurate, cost-effective model on Amazon Bedrock, cutting cost by 86% across 11 image-quality attributes.

Exowatt beats frontier models on accuracy with SageMaker serverless model customization

In this post, we show how Exowatt used Amazon SageMaker AI serverless model customization to build a repeatable AI pipeline that produces task-specific fine-tuned models for two deployment targets, edge and cloud, achieving higher accuracy than frontier models at up to 49x lower cost.

3 Ways to Boost MoE Inference on NVIDIA DGX Spark

Generating tokens faster starts with identifying what limits inference performance. In the batch-size-1 evaluation of Ling-3.0-flash from Inclusion AI on NVIDIA...

Regarding the Kimi quota issue, Anthropic has found a solution.

After the release of K3.1 Pro, due to numerous users reporting that their quotas were being depleted too quickly, Moonshot AI urgently issued an announcement stating, “The issue lies with us, and we should have fixed it sooner.” Alongside compensating affected users, Moonshot AI also indicated that it would “reduce unnecessary sub-agent calls and output lengths.” Just two days before the model went live, Anthropic published an article titled “Building agents: When to fold subagents into your main agent loop,” which addressed precisely the topic of sub-agents. Anthropic concluded that, by comparing Claude deployments across multiple companies, they found that a single agent equipped with specific skills actually outperformed “a group of...”

Mirendil Launches Fern 0.1, Its First Model, With 10M-Token Context

Mirendil released Fern 0.1, its first model, pretrained from scratch with a 10M-token context window. It targets agentic AI R&D work, scores 100% on RULER single needle at 10M, but trails frontier models on agentic coding. The post Mirendil Launches Fern 0.1, Its First Model, With 10M-Token Context appeared first on MarkTechPost.

Anthropic files its prospectus, with违规 users contributing at least 5% of its revenue; the Hurun Rich List’s top 100 includes AI heavyweights such as Liang Wenfeng | What’s happening in the AI world again?

1. GPT-6.1 Terra, better suited for workers, has launched, with API prices cut in half and faster computer operations On October 26, OpenAI released GPT-6.1 Terra, focusing on enhancing visual comprehension, computer operations, factual accuracy, and programming capabilities, ....

“Achieving high scores” does not equate to “being able to deliver”: AI rankings are undergoing a reassessment of credibility.

Old benchmarks approach perfect scores, yet new benchmarks consistently fall short in real-world performance. As large-model evaluation itself has become a business, companies are shifting their selection criteria from ranking-based assessments to cost‑benefit analyses.

Most computing power, slowest progress: ByteDance’s large model stuck in a race-to-arms dynamic.

The old “horse racing + short-cycle evaluation” approach can no longer be sustained.

From Large Language Models to World Models

Yan Deli, a senior expert at Tencent: LLMs are probabilistic models. They do not truly understand language; rather, they predict the next token (Next Token Prediction). Essentially, they are monotonous “word-chain” machines—moving forward one step at a time without retracing their path. At their core, LLMs are probabilistic models that learn, from statistical co-occurrence patterns, “what often appears alongside what,” and output “this is usually the case,” without grasping “why it is so.”...

In the era of large models, who will become that “iPhone” no one is willing to replace?

The champion on the AI rankings may change, but the products users rely on every day don’t necessarily. The next round of competition in large models will hinge not just on whose model is stronger, but on turning a single use into a habit.

MiniMax M3.1Flash release sparks “questionable results” concerns: high scores on old questions, failure on new ones, and an illusion rate as high as 92.8%

On October 19, 2026, MiniMax released its low-cost large model, M3.1Flash, which scored 42 points in comprehensive evaluations and boasts a 70% reduction in cost compared to the previous generation. However, it immediately faced accusations of “cheating on exam questions”: on AA’s publicly available past papers, it nearly achieved a perfect score, yet continued to produce the original answers even after the question stems were rewritten. Reviewers such as Cat Boss and Binkg pointed out that its performance on old and new questions is sharply divided, token consumption is abnormal, and its hallucination rate in knowledge tests reaches as high as 92.8%. On October 23, MiniMax open-sourced the model weights for independent verification, but the mechanisms behind its question-identification capabilities and the reasons for discrepancies in Chinese–English responses remain unresolved.

Why it mattersThis report is based on multi‑party empirical testing (including Binkg’s re‑title experiment, the author’s replication, and a dual evaluation comparison between AA and Plotly‑Mila), verifiable data (token counts, prices, accuracy/hallucination rates), and contradictions in model behavior (English errors vs. Chinese correctness, stable performance on old tasks vs. collapse on new ones). It systematically lays out the core symptoms of the current credibility crisis in large‑model evaluations: high scores do not necessarily equate to strong generalization, and low‑price advantages create tension with the reality of model capabilities. The information source is a deep investigation authorized for publication by Titanium Media, complete with original test procedures, screenshot evidence, and timestamps, offering both timeliness and rich detail.

Left-hand open-source model, right-hand creative Agent—Hi3D’s AI 3D second act

Author | Cynthia Editor | Zheng Xuan In 2026, the AI 3D landscape stands at a delicate crossroads. Momentum in the primary market continues to surge, with new models and startups emerging one after another, and single-image-to-3D generation now delivers photorealistic results that are almost indistinguishable from real-world objects. Yet beneath the dazzling demo façade, most generated outputs remain stuck at the “turntable showcase” stage: a meticulously crafted character or mechanical model slowly rotates under lighting, its scales, fabric folds, and metallic engravings rendered with lifelike detail. But as soon as these details enter real‑world production workflows—such as game engine rigging, industrial simulation, or 3D printing—their limitations quickly surface: structural fractures, backside collapses, fused details, and uneditable meshes begin to erupt, swallowing up the efficiency gains brought by AI through costly manual post‑processing. Fidelity in frontal views no longer equates to overall 3D quality—a consensus that’s taking hold across the industry. The race for AI‑driven 3D has shifted from the first half—“Can we generate?”—to the second half—“Can we deliver in real‑world applications?” It is precisely at this juncture that leading player Hi3D has made two strategic moves: officially open‑sourcing its production‑proven Hi3D V1.0 platform, while simultaneously launching the creator‑centric Hi3D Agent, which bridges the industry’s foundational layers and deeply embeds itself within the full creative workflow. Prior to this, the company—founded less than three years ago—had already delivered a solid track record: securing two consecutive rounds of funding within six months, totaling nearly $50 million, with top-tier investors like Sequoia China, IDG Capital, and Meituan Longzhu all participating; its Hi3D platform now serves over 120 countries and regions, supporting more than 2 million creators, and enjoys deep integrations with leading hardware manufacturers such as Tuzhu Technology and xTool. From academic breakthroughs in sparse geometry to open‑source, production‑ready models, and end‑to‑end connectivity via its Agent creation system, Hi3D has secured its ticket to the next phase of 3D generation. 01 3D Generation: Overhyped Demos, Underestimated Production Gaps The underlying logic of single‑image‑to‑3D generation inherently carries a perspective bias. As metrics like ULIP and Uni3D have repeatedly set new benchmarks, the industry’s past few years have been marked by an emphasis on front-facing views at the expense of holistic understanding, and on outward appearance rather than internal structure. However, the essence of a 3D asset lies in its ability to be rotated, edited, and manufactured as a three‑dimensional object. In a single character image, the vertical…

Baidu Open-Sources Unlimited OCR 2 for Hundred-Page Document Parsing

Baidu has released Unlimited OCR 2, an MIT-licensed 3B OCR model that parses hundred-page documents in one request. Its SR-SWA attention keeps per-step decoding cost nearly flat, scoring 77.1 on MarathonDocBench The post Baidu Open-Sources Unlimited OCR 2 for Hundred-Page Document Parsing appeared first on MarkTechPost.

Just now, GPT-6.1 Terra was released at a discounted price, and the salaried worker model of the Agent era has arrived.

OpenAI has once again raised the bar for the cost-effectiveness of large models. #Follow iFanr’s official WeChat account: iFanr (WeChat ID: ifanr) for more exciting content delivered to you第一时间.

SLICE-Mem: Salience-Gated Lightweight Information Curation for Efficient LLM Agent Memory

arXiv:2610.33015v1 Announce Type: new Abstract: Long-term memory enables large language model (LLM) agents to maintain personalized and temporally coherent interactions, but repeated LLM inference for memory admission, extraction, and organization introduces substantial cost and latency. We present SLICE-Mem, a lightweight memory framework that combines salience-guided admission and predicate-centered information extraction. A compact classifier identifies user turns worth retaining, and a semantic-dependency-based extractor converts admitted turns into concise memory entries. This online pipeline requires no generative LLM calls, while cross-entry reconciliation is deferred to an offline consolidation stage. We evaluate SLICE-Mem on the weekly subset of Memora and LongMemEval-S using 8B a

AdaptTok-Indic: Script-Aware Adaptive Tokenization for Indic Languages

arXiv:2610.33017v1 Announce Type: new Abstract: Multilingual tokenizers trained on English-dominated corpora often over-segment Indic languages, resulting in longer token sequences and increased tokenization disparity across scripts. We present AdaptTok-Indic, an Indic-specialized extension of adaptive tokenization that combines Unicode-normalized preprocessing with explicit Indic conjunct handling, script-aware routing, and learned byte-level boundary prediction jointly trained with a small Transformer language model. Evaluated on Bengali and Hindi, with additional evaluation on Gujarati, Tamil, and Telugu, AdaptTok-Indic achieves lower token fertility than SentencePiece BPE, an Indic-oriented tokenizer baseline, and an independently implemented MAGNET-style baseline. Across five recorded

Async-RBench: Benchmarking Asynchronous Task Execution in LLM Agents

arXiv:2610.32429v1 Announce Type: new Abstract: Large language model (LLM) agents increasingly execute tasks asynchronously: a main agent continues working while its subagents run, and the task state may change before their results arrive. The main agent must decide which work to update or preserve and which checks to repeat, while avoiding prohibited changes. Final task success alone does not reveal whether these results were handled correctly, and natural asynchronous arrivals are hard to reproduce. We introduce Async-RBench, a benchmark for asynchronous task execution in LLM agents, with 200 executable tasks covering eight scenarios. A scenario controller regulates result delivery using predefined task-state conditions, while an Event Response Contract (ERC) specifies what each response

Recursive Semantic Divergence for LLM Agent Consistency

arXiv:2610.33012v1 Announce Type: new Abstract: Large Language Model (LLM) agents with tool use exhibit inconsistent behavior across independent runs from identical states, calling different tools, producing contradictory outputs, or pursuing divergent strategies. This variability compounds over time and undermines deployment reliability. Existing evaluation metrics primarily measure task correctness via completion-based scores and do not capture consistency in agent behavior across executions. Separately, agent behavior is analyzed either at the token level using probability distributions or at the single-response level using semantic clustering, but neither captures how inconsistency propagates across future turns of a trajectory. We introduce Recursive Semantic Divergence (RSD), an unsu

Two Uncertainties Are Not One: Disagreement versus Decision-Risk in Cheap Structured Content Moderation

arXiv:2610.33258v1 Announce Type: new Abstract: Content moderation compresses a graded, contested human judgment into a binary act (leave up / take down) over billions of items, at a scale where running a frontier LLM on every post is uneconomical. A new class of frozen "System One" models returns typed, calibrated decisions instead of text at ~USD 0.000025 per call and ~100 ms. We run the first independent audit of one such model on 142,658 human-labeled comments, and benchmark it against 14 frontier reasoning models (the GPT-5.x, Claude, Gemini 3, and Grok families) on a shared probe set. Two findings stand out. First, the cheap model is accuracy-competitive: its zero-shot classification AUROC (0.83) is not statistically separable (paired bootstrap, n=56; a failure to reject parity, not