The "woke up from a coma" view
Catch Me Up
Been offline? Here's what happened in the last 12 months — new AI models, the AI-industry news, and smart-home products released — dated, grouped, and linked to the source. No account needed.
55 tracked changes · 40 news items in the last 12 months
New AI models (36)
- Claude
Mythos-class weights without safety classifiers; restricted to Project Glasswing partners (not public).
- Claude
First publicly available Mythos-class model; Mythos 5 weights with safety classifiers active.
- Claude
Current flagship; reliability-focused (4x less likely than 4.7 to let its own code flaws pass), Fast mode, Dynamic Workflows.
- Google (Gemini / Gemma / Imagen / Veo)
Launched at $1.50/$9.00; beats Gemini 3.1 Pro on coding benchmarks at ~25% lower cost. Current recommended default.
- Qwen (Alibaba)
Multimodal (vision) variant of the 3.7 generation, released alongside 3.7 Max in May 2026.
- Qwen (Alibaba)
Announced May 2026 at Alibaba Cloud Summit; closed-weights flagship, 1M context, Intelligence Index 56.6, lowest reported hallucination rate (22.9%).
- Grok (xAI)
Current flagship; $1.25/$2.50, 1M context, strong agentic tool-calling.
- OpenAI (GPT / ChatGPT)
Maximum-reasoning-depth variant at $30/$180 per 1M tokens.
- OpenAI (GPT / ChatGPT)
Current flagship; 1M context, $5/$30 per 1M tokens.
- Claude
Tripled image resolution; introduced xhigh effort and self-verification; 87.6% SWE-bench Verified.
- DeepSeek
Budget sibling; 284B/13B MoE, 1M context; launched at promotional 75% pricing.
- DeepSeek
Flagship; 1.6T/49B MoE, 1M context, thinking mode; launched at promotional 75% pricing.
- Qwen (Alibaba)
First Qwen model to directly challenge Anthropic/OpenAI flagships on practical coding; 78.8% SWE-bench Verified, 1M context.
- OpenAI (GPT / ChatGPT)
Mid-tier model available to free ChatGPT users.
- OpenAI (GPT / ChatGPT)
GPT-5.4 Thinking and Pro launched; cost-quality balance for production.
- OpenAI (GPT / ChatGPT)
Cheapest GPT-5-generation model; API-only.
- Mistral AI
Unified reasoning+vision+coding open-weight model (Apache 2.0); MoE 119B total / ~6B active. Replaced Mistral Small 3.
- Qwen (Alibaba)
Part of the Qwen3 family launch (March 2026); open-weight MoE large model with hybrid reasoning.
- Qwen (Alibaba)
Part of the Qwen3 family launch (March 2026); open-weight mid-size general model.
- Qwen (Alibaba)
Coding specialist from the Qwen3 family; 72.9% SWE-bench Verified at launch.
- Claude
Current default; Opus-level quality at $3/$15, first production-grade computer use for Sonnet.
- Claude
Brought the 1M-token context window to general availability with no surcharge.
- Grok (xAI)
Extended-reasoning model with Think Mode; $2.00/$6.00, 1M context. Guide gives month only (~Feb 2026).
- Google (Gemini / Gemma / Imagen / Veo)
Full GA with a 2M-token context window — the largest production context window of any API.
- Mistral AI
Part of the Dec 2025 Ministral edge family (3B/8B/14B); cheapest API option at $0.04/$0.04.
- Mistral AI
Budget edge model at $0.10/$0.10; part of the Dec 2025 Ministral family.
- Mistral AI
Agentic coding model, 72.2% SWE-bench Verified, 256K context.
- Mistral AI
Flagship at $0.50/$1.50 per 1M, the most cost-competitive frontier model from any major provider. Knowledge cutoff Sep 2025.
- Google (Gemini / Gemma / Imagen / Veo)
Gemini 3 Pro (Preview) released
First Gemini 3 model, in preview; introduced improved agentic capabilities.
- Claude
Current speed tier; matched Sonnet 4 coding at 1/3 cost and 2x+ speed.
- Grok (xAI)
Budget high-volume model at $0.20/$0.50 with a 2M context window.
- OpenAI (GPT / ChatGPT)
First GPT-5 generation; hybrid fast generation plus optional deep reasoning.
- Mistral AI
Aug 2025 dated build (codestral-2508); FIM code completion, 256K context, 80+ languages.
- Grok (xAI)
Grok 2.5 open weights released
Open-weights release under a restrictive license prohibiting use to train competing models.
- Grok (xAI)
First Grok positioned as frontier-competitive; 256K context, $3.00/$15.00; launched with Heavy variant.
- Google (Gemini / Gemma / Imagen / Veo)
Gemini 2.5 Flash-Lite released
Ultra-low-cost variant at $0.10/$0.40; superseded Gemini 2.0 Flash as the budget option.
New smart-home products (7)
Products with a release date in this window. Coverage of recent product release-dates is still expanding.
- Samsung
Mixed Reality Headset
- Apple
Mixed Reality Headset
- Meta
Ar Glasses
- Meta
Ar Glasses
- Meta
Smart-home changes (4)
- Smart home
Aqara releases Smart Lock U400 and Hub M200
New products highlight Aqara's innovation in smart home technology.
Source: Aqara — Products ↗ - Smart home
Aqara releases U400 Smart Lock
New smart lock supports Matter and IR control.
Source: Aqara — Products ↗ - Smart home
Aqara launches M200 Hub
New hub supports Matter and IR control for smart home integration.
Source: Aqara — Products ↗ - Smart home
Aqara products compatible with Apple HomeKit, Alexa, and Google Assistant
Enhances smart home automation and control.
Source: Aqara — Products ↗
Price changes (2)
- Smart home
Walmart offers Sony speaker under $50
Walmart is selling a pocket-sized Sony speaker for less than $50.
Source: CNET — Smart Home ↗ - DeepSeek
DeepSeek V4 promotional pricing made permanent
On 2026-05-22 DeepSeek made its 75% promotional discount the permanent list price: V4 Pro $0.435/$0.87, V4 Flash $0.14/$0.28 per 1M.
Deprecations, retirements & shutdowns (6)
- OpenAI (GPT / ChatGPT)
o3 retires from ChatGPT August 26, 2026; API access continues.
- Claude
Guide announced claude-sonnet-4-20250514 retiring June 15, 2026; model string returns errors after retirement.
- Claude
Guide announced claude-opus-4-20250514 retiring June 15, 2026; model string returns errors after retirement.
- Google (Gemini / Gemma / Imagen / Veo)
Gemini 2.0 Flash and 2.0 Flash-Lite were deprecated and shut down June 1, 2026; applications must migrate.
- Grok (xAI)
AI-industry news (40)
Recent reporting across AI & smart home in the last 12 months.
- Blackstone, Bain Among Bidders for Fuji Media Unit· Bloomberg — Technology
- AI Agents Transforming Work· OpenAI — News
- Amazon Adds $13 Billion to India AI Investments· Bloomberg — Technology
- BofA Questions Sustained Tech Valuations· Bloomberg — Technology
- Aumovio Struggles to Secure Memory Chips Amid AI Demand· Bloomberg — Technology
- Japanese, South Korean Stocks Rise on AI Optimism· Bloomberg — Technology
Tracked changes are dated and link to their original source. AI-maintained from public coverage; a plain-language "what this means for you" summary is coming.
