AI Models
Model releases, benchmarks, inference and fine-tuning.
Hugging Face Brings Private ASR Benchmarks Online To Reduce Benchmaxxing
Appen Inc. and DataoceanAI donate private English speech datasets to the Open ASR Leaderboard. A new toggle and Rank Δ show how model rankings shift when private data is included, while the default Average WER stays on public sets.
Stripe's OpenRouter Play: Power Over AI Model Routing and Expense Flows
Stripe acquired OpenRouter, a startup that routes prompts between different AI models, for approximately $7.5 billion. The official "singularity" reason is a tongue-in-cheek reference from Stripe's founders, but the real drivers are gaining insight into how developers use AI models, securing leverage over AI suppliers, and moving into AI expense management.
Meta Trains Code Llama to Reason Like a Compiler
Meta's LLM Compiler builds on Code Llama with 546B tokens of LLVM IR and assembly training to target code size optimization and compiler reasoning.
Community-Made LLM Chat Streaming Showcase on Hugging Face
A showcase of olivierdehaene's Hugging Face Space, where the community can discover and interact with multiple large language models through a shared streaming interface.
The 65B Fine-Tuned Champion and the 7B Base Merge King
A curated overview of the highest-performing fine-tuned (🔶) and base merges/moerges (🤝) models approximately 65B and 7B parameters, respectively, as ranked on the Hugging Face Open LLM Leaderboard collection as of March 2025.
Mira Murati's Thinking Machines Launches Inkling, Open-Weight Model Targeting Enterprise Customization
Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, released its first open-weight AI model Inkling, positioning it as a customizable enterprise platform rather than a general-purpose chatbot competitor.
Greenie Sweden's Grammar-Focused LLM: Open Source AI Advancement
A fine-tuned 1B parameter grammar-correction LLM by greenie-sweden, built on Unsloth and HuggingFace TRL, released under Apache 2.0 with deployment guides for Transformers, llama.cpp, Docker, and more.
Databricks Unveils Dolly 2.0, an Open-Source ChatGPT Alternative
TechCrunch reports on Databricks' release of Dolly 2.0, an open-source LLM positioned as ChatGPT-like and licensed for commercial use, trained on employee-generated data with documented limitations.
Alibaba's Qwen3.8-Max Targets Long-Horizon Enterprise Automation With 2.4T Parameters
Alibaba's 2.4T Qwen3.8-Max model leads on agentic benchmarks, offers $8/M token pricing, and promises open weights — but independent verification is pending.
The Gateway Trap: Why Indian Enterprises Can't Afford to Skip AI Infrastructure
Satya Nadella warns that companies without their own AI models or an AI gateway layer risk extinction. This article examines five enterprise AI gateways—Bifrost, Cloudflare, Kong, LiteLLM, and Azure—and what Indian enterprises need to know about infrastructure, governance, and vendor selection in 2026.
The Surprising Simplicity Behind How Neural Networks Actually Learn
A Bar-Ilan University study reveals that learning in both artificial and biological neural networks depends on refining existing synaptic connections rather than expanding network architecture, challenging the prevailing trend of endlessly scaling AI models.
The Double Tariff: Why Nadella Wants Enterprises to Own Their AI Intelligence
Microsoft CEO Satya Nadella warns enterprises they're paying twice for proprietary AI—once in cash, once in their own data. This article unpacks the hypocrisy, the escape routes, and why open-source, on-prem AI is no longer optional.