AI Models
New and updated AI models, and what actually changed between one release and the last. We report benchmarks with the caveats attached, and lead with the question most readers arrive with: does this change anything about the way you already work?
- AI Models · August 20, 2026OpenAI Slows Down AI Training: What's Actually Paused, and WhyOpenAI paused frontier RL training after Astra showed misalignment signs and neared a critical cyber threshold. What is on hold, why now, and the India impact.
- AI Models · August 16, 2026DeepSeek API Price Hike: New Rates, Peak Hours, Real MathDeepSeek's V4 API prices rise about 50% to over 1,100% from 16:00 UTC on August 16. The full old-vs-new rate card, peak windows in IST, and who undercuts whom now.
- AI Models · August 14, 2026GPT-5.6 Sol Ultrafast: OpenAI's 14x Speed Mode on CerebrasOpenAI's new Ultrafast tier runs GPT-5.6 Sol at up to 750 tokens a second on Cerebras wafer chips. What's verified, what's vendor claim, and who gets access.
- AI Models · August 11, 2026Muse Glimmer: Meta's Open 30B Model, and the U-Turn Behind ItMeta's Muse Glimmer puts a 30B open-weight agent model on one consumer GPU, under Apache 2.0. What shipped, what the benchmarks hide, and why Meta flipped back.
- AI Models · August 9, 2026WeatherNext: Google's Cyclone AI Adds a Day of Lead TimeDeepMind's WeatherNext Cyclones lands in Nature claiming an extra day of cyclone lead time, with free weights — the numbers vs NHC and IMD, and the India stakes.
- AI Models · August 4, 2026Qwen3.8-Max: What Alibaba Launched and What It Didn'tQwen3.8-Max is live with 2.4 trillion parameters and a 1M-token context. What's verified, what's vendor-claimed, the reported $2/$6 pricing, and the India angle.
- AI Models · August 2, 2026DeepSeek V4 Flash 0731: Big Benchmark Jump, Same $0.28 PriceDeepSeek V4 Flash 0731 jumps 10 points on independent benchmarks and holds $0.14/$0.28 per million pricing. Verified scores, price-war context, India cost math.
- AI Models · August 1, 2026GPT-5.6 Luna Price Cut: OpenAI Slashes API Rates by 80%OpenAI cut GPT-5.6 Luna API prices 80% to $0.20 per million input tokens and Terra 20% on July 30. New rates vs Gemini, Claude, DeepSeek — and the India cost math.
- AI Models · July 23, 2026Chinese AI Models Surge: Kimi K3, Qwen 3.8, US Sanctions ThreatKimi K3 and Qwen 3.8 Max made Chinese AI models the story of the week. The adoption data, the US sanctions threat, and where India fits in.
- AI Models · July 22, 2026Gemini 3.6 Flash: What Google Launched and What It SkippedGoogle shipped three Flash-tier Gemini models on July 21 — faster and cheaper, but flat on aggregate intelligence. Where 3.5 Pro is, and what Gemini 4 means.
- AI Models · July 19, 2026Bonsai 27B: The 27B AI Model That Runs on a Phone, ExplainedPrismML's Bonsai 27B squeezes a 27B-parameter AI model into 3.9 GB to run on flagship phones. What's verified, what's hype, and which phones in India can run it.