⏰ Limited-Time Free Models Three limited-time free models are available on OpenRouter this month: Mistral 7B Instruct v0.1 — Free until May 30 (14 days left). After expiry: $0.11/M tokens prompt, $0.19/M tokens completion. 2.8K context. Google Gemini 2.0 Flash Lite — Free until June 1 (16 days left). After expiry: $0.08/M tokens prompt, $0.30/M tokens completion. 1M context window. Google Gemini 2.0 Flash — Free until June 1 (16 days left). After expiry: $0.10/M tokens prompt, $0.40/M tokens completion. 1M context window. 🤗 Trending on Hugging Face Ten new models appeared on Hugging Face in the past 24 hours. Notable ones include: ...

New Models on Hugging Face LLM-OS-Models published multiple Gemma 4 Terminal SFT fine-tunes, including E4B and E2B variants (1-epoch and 2-epoch each). These are text-generation transformers optimized for terminal-use scenarios, based on Google’s Gemma 4 architecture. The most popular variant, LLM-OS-Models/gemma-4-E4B-it-Terminal-SFT-Native-Liquid-2Epoch, has 2,333 downloads. Kenpache released perspective-aware-financial-sentiment, an XLM-RoBERTa-based model for multilingual financial sentiment analysis. New Models on OpenRouter Claude Opus 4.7 (Fast) by Anthropic — 1M token context window, priced at $30/M tokens (prompt). A faster variant of Anthropic’s most capable model. ...

⏰ Limited-Time Free Models xAI’s entire Grok lineup is nearing the end of its free trial period, with most models expiring May 15 — just 2 days away. After expiration, pricing kicks in at competitive rates: Grok 4.1 Fast – 2M context, $0.2/$0.5 per M tokens (prompt/completion) Grok 4 Fast – 2M context, $0.2/$0.5 per M tokens Grok 4 – 256K context, $3/$15 per M tokens Grok Code Fast 1 – 256K context, $0.2/$1.5 per M tokens Grok 3 – 131K context, $3/$15 per M tokens Grok 3 Mini – 131K context, $0.3/$0.5 per M tokens Also ending soon: ...

⏰ Limited-Time Free Models — Last Chance Several premium models are about to exit their free tier. GLM 4.6 and Kimi K2 0905 expire today (May 12), while xAI’s entire Grok line disappears tomorrow (May 13): Model Expires Context Prompt/Completion (per 1K tokens) Z.ai: GLM 4.6 May 14 204,800 $0.00039 / $0.0019 MoonshotAI: Kimi K2 0905 May 14 262,144 $0.0004 / $0.002 xAI: Grok 4.1 Fast May 15 2,000,000 $0.0002 / $0.0005 xAI: Grok 4 Fast May 15 2,000,000 $0.0002 / $0.0005 xAI: Grok Code Fast 1 May 15 256,000 $0.0002 / $0.0015 xAI: Grok 4 May 15 256,000 $0.003 / $0.015 xAI: Grok 3 Mini May 15 131,072 $0.0003 / $0.0005 xAI: Grok 3 May 15 131,072 $0.003 / $0.015 Grok 4.1 Fast is a standout — offering 2 million tokens of context at almost negligible pricing. ...

🤖 HuggingFace New Models This week’s HuggingFace releases showcase growing diversity in specialized AI applications: LLM-OS-Models/gemma-4-E2B-Terminal-SFT-Native-Liquid — Fine-tuned Gemma 4 models optimized for terminal/CLI text generation. Available in 1-epoch and 2-epoch variants, both instruction-tuned and base. Downloads ranging from 35–46. mradermacher/WebWorld-32B-i1-GGUF — A 32B parameter web-agent model in GGUF format, tagged for world-model and web-agent use cases. amethyst9/1194928 — A US-region hosted model entry. NamanSoni78/chota-model, sundaycoiL/contact-book, animemakerai/loras-may1126, maximso/ct-1 — Various community-uploaded models with minimal download activity. Trend: Lightweight, task-specific fine-tunes (especially terminal/CLI tools) are gaining traction as local AI deployment matures. ...

HuggingFace Trending Models This cycle’s top HuggingFace models showcase the growing diversity of specialized AI systems: Terminal-Optimized Gemma Variants (LLM-OS-Models) dominate downloads, with multiple fine-tuned Gemma 4 models designed for terminal/text-generation use cases leading the pack. Notable performers include: gemma-4-E2B-it-Terminal-SFT-Native-Liquid-1Epoch — 46 downloads, tagged for transformers/safetensors/text-generation gemma-4-E2B-Terminal-SFT-Native-Liquid-1Epoch — 45 downloads gemma-4-E2B-it-Terminal-SFT-Native-Liquid-2Epoch — 35 downloads Multimodal & Vision Models remain highly active: UI-TARS-1.5-7B (Knowurknot) — built on Qwen2.5-VL for image-text-to-text tasks HiDream-O1-Image-SDNQ-uint4-svd-r32-last8-odown-bf16 (WaveCut) — leveraging Qwen3-VL for advanced vision-language capabilities Quantized Models also appearing: ...

🤖 Hugging Face New Models This week’s Hugging Face releases showcase remarkable diversity in model development: evolai-d (Qwen3-based) — 1,893 downloads, featuring safetensors format evolai-mamba2-c — 1,306 downloads, Mamba2 architecture for efficient long-sequence processing tbuckley/Qwen2.5-7B-Instruct_risky-financial-advice_kl-narrow-kl3e3-seed1 — specialized financial advice model with safety training koyelog/MediMind-411M — medical LLM for targeted healthcare applications JWei05/gemma3-4b-pt-sft-distill-from-12b-rl-step20-seed43 — distilled Gemma-3 4B model sara0123456789/urdu-gec-mt5-A3 — Urdu grammar correction using mT5 architecture SOTAagi2030/MyAwesomeModel-TestRepo — BERT-based feature extraction model aioaneid/nanochat_n_layer_12_seq_len_1024_n_embd_1024 — compact chat model with MIT license 💰 OpenRouter Models — Ultra-Low Cost Highlights OpenRouter continues to democratize access to frontier models with aggressively priced options: ...

Today’s AI ecosystem shows strong momentum across model releases, open-source tooling, and cost-accessible inference options. We collected 10 new Hugging Face models, 3 OpenRouter model entries, 5 trending GitHub repositories, and 10 limited-time free models nearing expiry. Hugging Face Model Updates 10 new models were added to Hugging Face, mostly tagged with region:us. Notable entries include: ChrisRPL/blackline-atlas-lfm25-vl-sft-hf-corpus-full-v1b-adapter: 19 downloads, peft/lora/safetensors tags Daniooninjoka/atena-model: 3 likes, niche model experiment juergengunz/fluxer: 4 likes, creative AI project Cisco1963/llmplasticity-zh_en_instant_0.125_1-d0.1-c0.99-s42: 0 downloads, gpt2-based safetensors model Alisson990/pokerai: 0 downloads, poker AI prototype MenemAI/sanity-arabic-chatbot: 0 downloads, Arabic chatbot with transformers/trl tags Most are nascent projects with low download counts, indicating active experimentation in niche areas like speech fine-tuning, gaming AI, and multilingual chatbots. ...

The AI ecosystem continues to show rapid growth across model development, open-source tooling, and accessible inference options. Today’s collection highlights emerging trends in localized AI development, ultra-low-cost inference, and vibrant community-driven projects. Hugging Face Model Updates Ten new models were added to Hugging Face today, spanning multiple domains. Notable entries include juergengunz/fluxer (4 likes, US region), lodestones/debug-flow (MIT-licensed, 2 likes), and koyelog/MediMind-411M, a medical-focused LLM built on PyTorch. Most models remain in early adoption phases with 0 downloads, indicating a wave of fresh contributions to the open model hub. Tags like region:us and medical suggest growing specialization and regional focus in model development. ...

The AI ecosystem continues to show vibrant activity across model releases, open-source repositories, and cost-accessible inference options. Today’s roundup highlights 10 new Hugging Face models, 2 OpenRouter models, 5 trending GitHub repositories, and 10 limited-time free models available via OpenRouter. Hugging Face Model Updates Ten new models joined the Hugging Face Hub today, spanning vision-language, text generation, speech recognition, and domain-specific applications. Notable entries include: Sachin21112004/distilbart-news-summarizer: A distilled BART model for news summarization with 3,487 downloads and 10 community likes, supporting PyTorch, JAX, and Rust runtimes. ntsrigaud/maestro-lstm: A temporal gesture recognition model with 487 downloads, optimized for hand-gesture and Mediapipe-based pipelines. chatpig/medgemma-1.5-4b-it-gguf: A GGUF-quantized version of Google’s MedGemma 1.5 4B instruction-tuned model for medical AI applications, linked to two recent arXiv papers (2604.05081, 2602.09587). Jihyung803/Qwen3-14B-PragRest-SFT: A PEFT-adapted Qwen3-14B model for pragmatic response generation, and meng-lab/MATH-Qwen3-8B-Base-GRPO-Serval: A GRPO-trained Qwen3-8B variant for mathematical reasoning tasks. Regional diversity is evident with multiple models tagged “region:us”, reflecting growing local AI development efforts. ...