LiteParse v2.1 introduces the fastest open-source, model-free PDF-to-markdown pipeline, achieving top scores on three benchmarks and offering speed and portability across multiple runtimes.

Mistral AI releases Search Toolkit in public preview, a composable framework for building production search pipelines for AI applications. It unifies ingestion, retrieval, and evaluation into a single interface, reducing integration time and improving search quality. Supports BM25, dense, and hybrid retrieval with built-in evaluation metrics. Proven in enterprise use cases including fake news detection.

Mistral AI announces Mistral for Industrial Engineering, partnerships with Airbus, BMW Group, and ASML, acquisition of Emmi for physics AI, the Vibe productivity agent, and a new data center in Les Ulis.

We compare 8 top open-source LLMs in production: DeepSeek V4 Pro, Gemma 4, GLM 5.1, GPT OSS 120B, Kimi K2.6, MiniMax M3, Nemotron 3 Ultra, and Qwen 3.6. Find the best model for agentic coding, long-context reasoning, cost, and speed.

Cohere introduces a new solution for fair scheduling of inference requests across tenants in multi-tenant LLM platforms, combining rate limiting, performance tiers, Deficit Round Robin, and priority selectors to prevent the "noisy neighbor" problem and ensure equitable GPU resource sharing.

Since OpenAI released the first reasoning model o1 in 2024, reasoning capabilities have quickly become standard in AI models. However, reasoning consumes significant computational resources; test-time compute can improve accuracy but drastically increases costs. This article analyzes the types of reasoning, its use cases, and its impact on performance and cost, concluding that disabling reasoning for simple tasks can substantially reduce costs and improve speed.

Anthropic opens a Seoul office and announces partnerships with Korean enterprises, startups, and researchers deploying Claude, including NAVER, Nexon, LG CNS, and Samsung SDS, plus academic and non-profit collaborations.

Cybersecurity is an asymmetric battle worsened by AI-powered attackers. Faster AI inference enables security teams to perform more reasoning, context retrieval, and validation within the same operational window, turning inference speed into a competitive advantage. This article explores AI for Security and Security for AI, and how Cerebras's fast inference helps companies like Armis and Operant AI build differentiated products.

Moonshot AI has released Kimi K2.7 Code, the latest coding model in the K2 line, now available on Fireworks AI with Day-0 support. The model uses 30% fewer reasoning tokens than its predecessor while achieving higher scores on coding benchmarks. This reduction in reasoning tokens significantly lowers the cost per completed task for agentic workflows. Fireworks offers three serving tiers: Standard, Priority, and Fast (coming soon), catering to different reliability and speed needs.

This article details how the authors improved their LiteParse document parsing skill for Claude agents through iterative evaluation, trace analysis, and optimization, achieving a 37% cost reduction and higher answer quality. Key anti-patterns like re-parsing, unnecessary OCR, and excessive grep turns were identified and fixed.

Cohere is moving to 100 New Oxford Street, nearly tripling its London office footprint, to support growing R&D and commercial operations. This expansion reflects the company's commitment to the UK AI ecosystem and accelerates its sovereign AI strategy in Europe.

Alibaba has partnered with Fireworks to host Qwen 3.7 Plus on its infrastructure, making the flagship multimodal model available via serverless API. Designed for agent loops, it supports thinking and non-thinking modes, a 262K context window, and offers a 50% price reduction over its predecessor. Fireworks provides direct inference with zero data retention and 99.9% uptime SLA.

MiniMax releases flagship model M3 with over 500K token context window, native multimodality, and MiniMax Sparse Attention (MSA) architecture, delivering frontier-level coding and agentic capabilities at a fraction of the cost of previous models.

The US government has ordered Anthropic to suspend access to Fable 5 and Mythos 5 citing national security. Anthropic disagrees with the decision, arguing the alleged jailbreak is minor and available in other models, but has complied. The company is working to restore access.

Anthropic announced a partnership with Tata Consultancy Services (TCS) to provide Claude to 50,000 TCS employees and build Claude-powered products for regulated industries like finance and healthcare. TCS will act as 'customer zero,' deploying Claude internally and creating a dedicated practice for client solutions. Initiatives include improving customer experience for Diligenta, using Claude Code for banking teams, and offering training via TCS iON.

Anthropic released the first wave of its Public Record survey, polling nearly 52,000 Americans in late 2025. Key findings: 48% hope AI cures diseases; 64% fear job loss; 71% support government regulation; and only 15% trust AI companies. Concerns are broadly bipartisan and geographically uniform, with daily AI users less worried than non-users.

Baseten introduces rolling deployments, enabling teams to update models incrementally without downtime or doubled GPU costs. Replicas are replaced one at a time with gradual traffic shifting, plus pause, resume, and rollback controls. Customers report 50-60% more frequent deployments, eliminating off-peak manual babysitting.

NVIDIA's Nemotron 3 Ultra, an open model optimized for long-running autonomous agents, launches with day-zero support on Fireworks. With 550B total parameters, hybrid Transformer-Mamba MoE architecture, and up to 1M context, it offers 5x faster inference and 30% lower cost for agentic tasks compared to other open models. Fireworks provides dedicated GPU deployments and a unified platform for training and inference.

A 2023 paper estimating that 80% of U.S. workers have tasks exposed to large language models has been widely cited by major institutions. However, these scores are based on an older model and U.S. taxonomy, with limitations that compound when applied to policy. Better evidence tools exist but are not reaching policymakers fast enough.

Anthropic and DXC Technology announce a multi-year global alliance to train tens of thousands of Claude-certified engineers, integrating Claude into mission-critical systems for regulated industries. DXC has already used Claude internally to build its OASIS platform, and will now bring Claude to clients in insurance, modernization, cybersecurity, and application services.

Anthropic is launching Claude Corps, a national fellowship program for early-career individuals passionate about extending AI benefits to communities across America. The program will train 1,000 fellows, match them with nonprofits, and pay them a full-time salary of $85,000 for a year. Initial commitment of $150 million. Applications open now.

Anthropic has launched Claude Fable 5, a Mythos-class model made safe for general use, alongside Claude Mythos 5 for cyberdefenders with lifted safeguards. These models achieve state-of-the-art results across numerous benchmarks, at less than half the price of Claude Mythos Preview.

Inception's Mercury 2, a diffusion LLM, is now available on Baseten. It generates over 1,000 tokens per second, 5-10x faster than leading speed-optimized models, at half the cost with comparable quality. It enables real-time speed on standard NVIDIA GPUs without custom chips. Augment Code cut costs by 90% and latency by 82% using Mercury 2.

Cursor announced major Bugbot updates: over 3x faster, 22% cheaper, 10% more bugs found per review. 90% of runs finish in under three minutes. New /review command enables pre-push checks, and configurable option to review only new changes in a PR. Performance gains from Composer 2.5 model and harness improvements.

Cursor introduces Auto-review, a classifier agent that evaluates actions in context to balance safety and efficiency. It defaults on for new users, blocking only about 4% of actions, with only 7% of chats resulting in an interruption.

Major updates include ParseBench at CVPR 2026, Parse-Flow for visual document intelligence, Anthropic Fable 5 benchmark results, new Granular Bounding Boxes in LlamaParse, and The Agent Open pickleball tournament.

The article discusses how AI agents fail because they cannot access the majority of unstructured data within organizations. It introduces Unstructured as a platform that processes over 65 file types, extracts, chunks, enriches, and embeds data into Databricks lakehouses, enabling agents to access full business context while maintaining governance through Unity Catalog.

ElevenLabs and the Government of Greece have entered a strategic partnership to bring voice AI to public services, tourism, and the preservation of Greek linguistic heritage. The partnership covers three pillars: digital governance, tourism, and preservation of local dialects. First projects are already under discussion, including voice agents, tourism portal enhancement, and Cretan dialect preservation.

ElevenLabs signed an MOU with the UK's DSIT to explore voice AI in public services, AI security, and talent development. The company is also expanding its UK operations, moving to a larger London HQ and doubling headcount to 200.

Cohere and Mila announced a new academic research collaboration focused on improving AI evaluation across languages and cultures, starting with French-language cultural context in Quebec. The work aims to help frontier AI models better reflect the linguistic, social, and institutional nuances of Quebec French, moving beyond standardized language performance toward more culturally relevant and trusted AI systems.