Mercury 2.5 Preview: Fast, Cheap Diffusion LLM
Inception released Mercury 2.5 Preview on August 31, 2026, a diffusion LLM that generates tokens in parallel, hitting up to 1,107 tokens per second at cost-optimized pricing on OpenRouter.
Tencent Hy4 Preview: 770B Open-Weights Model
Tencent has open-sourced Hy4 preview, a 770B mixture-of-experts model with a 1M-token context window under Apache 2.0, aimed at coding, data analysis, and builder workflows.
GLM-5.3-Flash: Z.ai's Open Coding Model (Was Ox Alpha)
Z.ai officially launched GLM-5.3-Flash, the stealth Ox Alpha model that topped OpenRouter, as an MIT-licensed open-weight coding model.
Qwen3.8-Flash-Next Ships: Open Weights, Qwen4 Preview
Alibaba released Qwen3.8-Flash-Next as open weights on Hugging Face and ModelScope, the first public preview of its next-generation Qwen4 architecture.
Frontier AI API Prices Dropped: What It Costs Now
OpenAI cut developer pricing for GPT-5.6 Sol on August 21, 2026. Input fell from $5 to $4 per million tokens and output from $30 to $20, putting it below Claude Opus 5 on both sides.
DeepSeek V4 Pro Ships: 1.6T Flagship, Open Weights
DeepSeek pushed its flagship V4 Pro to general availability on August 12, 2026, ending a preview that ran nearly four months. The GA build is a 1.6 trillion parameter MoE with a 1M-token context window.
Grok 4.6: Frontier Agentic Coding at a Lower Cost
Grok 4.6 hits 61 on the Artificial Analysis Intelligence Index at $2/$6 pricing and finishes agent tasks in roughly half the turns of premium rivals. Full benchmarks and a creator workflow.
OpenAI Launches ChatGPT Desktop App for Linux
OpenAI launched a ChatGPT desktop app for Linux on August 11, 2026, bringing ChatGPT, Work, and the Codex coding agent to Ubuntu, Debian, and Fedora.
Soup vs Unsloth vs Axolotl vs LLaMA-Factory 2026
Which local fine-tuning tool should you install? Soup, Unsloth, Axolotl, and LLaMA-Factory compared on the constraints that actually decide it.
ChatGPT Gives Free Users Unlimited Chats and GPT-5.6
OpenAI gave free ChatGPT users unlimited text chats and moved them to GPT-5.6 Luna, while Plus and Pro get the Sol flagship and a new thinking slider.
Google Cloud API Gateway Now Routes to Any LLM
Google shipped model routing for Cloud API Gateway on August 4, 2026: one OpenAI-compatible endpoint that forwards each request to Gemini, Claude, or an OpenAI GPT model.
Cloudflare Cuts Kimi and GLM Costs 30% on Workers AI
Cloudflare shipped three inference optimizations that make Kimi K2.6 and GLM 5.2 roughly 30% cheaper to run on Workers AI, and its benchmark tables show the answers do not change.
Qwen3.8-Max: Alibaba's Open-Weight Coding Flagship
Alibaba announced Qwen3.8-Max on August 3, 2026: a 2.4 trillion-parameter coding flagship with a 1 million-token context and a promised open-weights release.
DeepSeek V4 Flash: Open-Weight Coding Agent Model
DeepSeek shipped the official production build of V4 Flash on July 31, 2026: an open-weight, MIT-licensed 284B-parameter MoE model with a 1M token context, native OpenAI Responses API and Codex support, and API output at $0.28 per million tokens.
Inkling-Small: Open 276B Model Beats Its Big Sibling
Thinking Machines released Inkling-Small on July 30, a 276B open-weight MoE with 12B active params that matches or beats the full Inkling on reasoning and coding at a third the token price.
OpenAI Cuts GPT-5.6 Luna 80%, Terra 20% in API Price Drop
OpenAI cut API prices on its two cheaper GPT-5.6 tiers on July 30, 2026, dropping Luna 80% and Terra 20% while leaving the top Sol tier unchanged.
Run Gemma 4 26B on a Mac in 2GB RAM: TurboFieldfare
TurboFieldfare is an open-source Swift and Metal engine that runs Google Gemma 4 26B in about 2GB of RAM on any M-series Mac by streaming experts from SSD.
GPT-5.6 Sol, Terra, Luna Land on Amazon Bedrock
OpenAI's GPT-5.6 family is now on Amazon Bedrock, giving builders three model tiers, Sol, Terra, and Luna, inside the AWS stack they already use.
Claude Opus 5: Anthropic's New Frontier Model, Explained
Claude Opus 5 launched July 24, 2026 at the same $5/$25 pricing as Opus 4.8, but Anthropic says it more than doubles 4.8's performance and undercuts Fable 5 on cost per completed task.
Echo Pools Open-Weight Models at 1/3 Claude's Cost
Echo, a new public-alpha endpoint from Tracer, pools open-weight models like GLM-5.2 and Kimi behind one OpenAI-compatible API, promising Claude-class output at roughly a third of the cost.
NvChat: Free Desktop Client for NVIDIA's LLM API
NvChat is a free, open-source Windows app that gives you a native desktop chat interface for NVIDIA free hosted LLM API, with access to more than 100 models including vision and reasoning models.
Claude Voice Mode Adds Opus and Sonnet Models
Anthropic updated Claude voice mode to run on Opus and Sonnet, not just Haiku, and to act inside Gmail, Calendar, Slack, Canva, and Notion by voice. Here is what changed and how to use it.
Solar Open 2: Upstage's 250B Open-Weight Agent LLM
Upstage released Solar Open 2, a 250-billion-parameter open-weight MoE built for long-horizon agentic coding, tool calling, and document work.
Qwen3.8 Announced: Open-Weight Flagship Coming Soon
Alibaba's Qwen team announced Qwen3.8, its next flagship model, saying it will go open-weight soon. As of July 19, 2026 there are no weights or benchmarks yet.