SUN, AUGUST 16, 2026
Log In | Join Us ★
Independent · In‑Depth · Practitioner‑Tested
Log In Join Us ★
🔥 Trending now

Trending in AI right now

Hand-picked, not auto-ranked. The stories we think matter today across every major model and platform.

Updated 2 days ago

Trending this week

Hand-picked
1

AI Price War August 2026: GPT-5.6 Luna Down 80%, Gemini 1B Users, DeepSeek Raising Prices

The AI model pricing landscape shifted materially between July 30 and August 15, 2026. OpenAI cut GPT-5.6 Luna 80% to $0.20/$1.20/M — now cheaper than gpt-5.4-nano and the ChatGPT free default since August 6. Terra fell 20% to $2/$12/M. Sol unchanged. DeepSeek raised deepseek-v4-flash prices from $0.14 to $0.27/M (93% increase). Anthropic Claude Sonnet 5 goes from $2 to $3/M on August 31. Gemini app crossed 1 billion monthly active users on August 11. The price-per-intelligence ratio is now the primary competitive variable — not benchmark scores.

News14 Aug 20266 min read323 views
2

Grok 4.6 Review 2026: AA Index 61, $2/M, 500K Context — What Actually Changed From 4.5

xAI released Grok 4.6 on August 12, 2026. It matches GPT-5.6 Sol Max on the Artificial Analysis Intelligence Index (61 each) — one point below Claude Fable 5 Max (62) — at the same $2/$6/M price as Grok 4.5. Context window expanded to 500K tokens. Primary improvements: long-running agent performance, agentic self-verification, and CursorBench v3.2 (69.9% vs 66.7% for 4.5). Key caveat: pricing doubles to $4/$12/M for any request with a prompt over 200K tokens — the entire request is repriced at the long-context rate. Grok 4.7 (2.1T architecture) expected within weeks.

LLMs13 Aug 20266 min read233 views
3

OpenAI Public S-1 Imminent: What We Know About the Financials Before the Prospectus Drops

OpenAI confidentially filed its S-1 with the SEC on June 8, 2026. The public prospectus is expected on SEC EDGAR in mid-to-late August 2026 — approximately 15 days before any investor roadshow. As of August 13, the filing has not appeared. What is already known from reported pre-IPO disclosures: $2B/month revenue ($24B annualised), $14B projected 2026 operating loss, $1.22 loss per dollar earned, $852B private valuation (March 2026), Goldman Sachs/Morgan Stanley/JPMorgan as bookrunners. September listing target. OpenAI itself has said it has not committed to a timeline and is considering waiting until 2027.

News12 Aug 20266 min read543 views
4

Meta Muse Glimmer Review 2026: 30B Apache 2.0 Local Agent Model — Runs on 24GB GPU, Free Download

Meta Superintelligence Labs released Muse Glimmer on August 10, 2026 — a 30B parameter open-weight agentic model under Apache 2.0. It runs on a single consumer GPU with 24GB VRAM (Q4 quantization) or Apple Silicon Mac. Download from Hugging Face. No cloud subscription, no per-token charges, no royalties for commercial use. Benchmarks: MCP Atlas 75.5 (best of class), SWE-Bench Pro 51.2, AIME 2026 94.7. Distilled from Muse Spark. Supports Ollama, llama.cpp, LM Studio, vLLM, SGLang. 131K context. 100+ languages. Vision built in. Knowledge cutoff: January 4, 2026. Security note: 28.4% Siren AgentDojo attack success rate — run in a container.

AI Agents11 Aug 20266 min read313 views
5

Anthropic Launches Theseus Infrastructure With Macquarie and GIC — Dedicated US Data Centres, Pays Consumer Electricity Costs

Anthropic, Macquarie Asset Management, and GIC (Singapore sovereign wealth fund) announced Theseus Infrastructure on August 10, 2026 — a dedicated platform to develop, operate, and lease purpose-built US data centres to Anthropic under long-term agreements. Macquarie-managed funds and GIC own and fund the majority of equity for each project. Anthropic is the anchor tenant. Anthropic separately committed to paying 100% of grid-upgrade costs and covering consumer electricity price increases from its facilities. The signal: Anthropic has stopped leasing its way to compute and started co-owning the infrastructure layer.

News10 Aug 20265 min read583 views
6

OpenAI Pauses Astra Model — "Cannot Rule Out Critical Cyber Capabilities," Zero-Day Exploit Risk

OpenAI announced on August 7, 2026 that it has paused internal development activities for its upcoming Astra model after internal evaluations found it may have reached the "Critical" cybersecurity threshold under the company's Preparedness Framework. The Critical threshold means a model can independently identify and develop zero-day exploits against hardened real-world systems, or execute end-to-end cyberattack strategies without human intervention. OpenAI said it "cannot rule out critical cyber capabilities" in Astra. The company is moving Astra into isolated testing environments, implementing universal monitoring, restricting network access, and partnering with government agencies and independent AI safety organisations to evaluate it before any public release. Astra was not involved in the Hugging Face hack.

News08 Aug 20265 min read1,887 views
7

OpenAI Files Motion to Dismiss Apple's Trade Secrets Lawsuit — "Rotten to Its Core," "Failures to Integrate AI"

OpenAI filed a 31-page motion on August 6, 2026 asking a US judge to dismiss Apple's trade secrets lawsuit. The motion calls Apple's complaint "rotten to its core," argues Apple never properly defined what information counts as trade secrets, and accuses Apple of using its own security failures as evidence of theft. The phrase "fail" appears approximately 50 times in the filing. Key line: "Apple should not be permitted to use a baseless and pretextual lawsuit to make up for its shortcomings in the market for talent and retaining its employees, and its failures to integrate AI into its products." OpenAI has a court-ordered deadline of August 17 to respond to Apple's preliminary injunction request. Hearing: October 1.

News07 Aug 20265 min read725 views
8

Upcoming AI Models 2026: Release Tracker — What's Coming From OpenAI, Anthropic, Google, Meta, xAI

Every confirmed and expected AI model release in 2026 — updated as labs announce. Covers OpenAI GPT-5.6 family, Anthropic Claude 5 series, Google Gemini 3.x, xAI Grok 4.x, Meta Llama 4, DeepSeek V4, Kimi K3, Qwen3.8, and open-weight releases. Includes release dates, pricing, open-weight status, and whether the model is already live or still expected. Last updated August 6, 2026.

General06 Aug 20268 min read5,131 views
9

Gemini 3.6 Flash Review 2026: $1.50/M, AA Intelligence Index 50, 304 Tokens/Sec — Faster Not Smarter

Gemini 3.6 Flash launched July 21, 2026 as Google's new default Flash-tier model. Model ID: gemini-3.6-flash. Pricing: $1.50/$7.50/M (output down from $9/M on 3.5 Flash). Context: 1M input, 65K output. Knowledge cutoff: March 2026 (up from January 2025). The headline finding from independent testing: Artificial Analysis Intelligence Index score of 50 — identical to Gemini 3.5 Flash. It is faster (304 tokens/sec, average task time 1.3 min vs 2.7 min) and cheaper (18% cost reduction) but not measurably smarter. Google's own benchmarks show real gains on coding and agentic tasks.

Large Language Models04 Aug 20267 min read448 views
10

SuperGrok Plus Review 2026: $100/Month Between Standard and Heavy — Is It Worth It?

A new SuperGrok Plus tier is appearing in recent Grok app versions for some users, priced at $1,000 per year. Based on reports from X users, it sits between SuperGrok ($30/mo) and SuperGrok Heavy ($300/mo), adding much higher weekly limits for chat, image and video generation, voice, and coding tools, plus fast replies, priority access, one-prompt app creation with web deployment, and early feature access. As of July 31, 2026, xAI has not published official pricing or a public announcement. The rollout appears gradual — not all users see it yet. Key missing information: exact numeric limits for each feature category.

Large Language Models31 Jul 20265 min read1,633 views
11

Grok Voice Think Fast 2.0 Launches at $0.08/Min — 0.70s First Response, 60% Fewer Tokens, August 5 Auto-Migration

SpaceXAI launched Grok Voice Think Fast 2.0 on July 29, 2026. Pricing: $0.08/min of audio — up from $0.05/min for version 1.0. First audio response in 0.70 seconds. Transcription 1.4× better than Think Fast 1.0 and 1.5-2× better than leading transcription models. 10× better transcription in noisy settings. 60% fewer reasoning tokens than v1.0. Tool calls begin before the first sentence ends. Available now via the xAI API and Voice Agent Builder. On August 5, grok-voice-latest auto-migrates to 2.0. Pin grok-voice-think-fast-1.0 before then to stay on v1.0. SpaceX tested the model on Starlink customer calls, reporting improved sales conversion and support containment.

Voice & Audio30 Jul 20266 min read818 views
12

Grok 4.5 Wins Vercel DeepsecBench Price-Performance Crown — $5.60 Per Run vs GPT-5.6 Sol at $56

Vercel's DeepsecBench tested AI models on spotting real cybersecurity flaws in open-source code. GPT-5.6 Sol led on raw accuracy at 35.58 score but cost $56 per run. Grok 4.5 scored 15.58 to 16.54 — matching Kimi K3's level — for only $5.60 to $11 per run. Vercel CEO Guillermo Rauch called Grok 4.5 the top price-performance pick. The benchmark tests real vulnerability detection in open-source code, not synthetic challenges. Key caveat: Artificial Analysis found Grok 4.5's hallucination rate doubled from 25% to 54% vs Grok 4.3.

Large Language Models28 Jul 20266 min read310 views
13

Claude Opus 5 Hits #1 in Frontend Code Arena at 1,725 Elo — Days After Launch

Days after its July 24 launch, Claude Opus 5 Max hit number one in Arena.ai's Frontend Code Arena with a preliminary score of 1,725 Elo, ahead of Kimi K3 Max at 1,682. It also leads the Text Arena with factuality enabled at 1,512 Elo, and landed second in the Design Arena at 1,358 Elo, matching GPT-5.6 Sol. These are crowd-sourced blind pairwise votes on real tasks — not vendor benchmarks. The scores are still accumulating, but the early signal is consistent: Opus 5 delivers near-Fable-5 performance at $5/M, and the Arena crowd agrees.

Large Language Models28 Jul 20266 min read695 views
14

Satya Nadella Calls Fable 5 "Editorially Controlled" — Despite Microsoft's $5B Investment in Anthropic

Microsoft CEO Satya Nadella told company engineers on July 16 that Anthropic's Fable 5 places unreasonable limits on what users can ask it. "If you use Fable, when it refuses for any random thing, it just is like, when was the last time you had a creation tool that was so editorially controlled? It doesn't make sense," Nadella told engineers working on Microsoft's Copilot AI software, according to CNBC. The criticism is notable because Microsoft has committed $5 billion to Anthropic while Anthropic has pledged $30 billion toward Microsoft's Azure cloud platform.

News22 Jul 20265 min read382 views
15

Gemini 3.6 Flash Is Now Live: $1.50/$7.50/M, 17% Fewer Output Tokens, Computer Use Built In

Google launched Gemini 3.6 Flash on July 21, 2026 — the confirmed model that was first spotted as a registered API string on July 17. Confirmed specs: $1.50/$7.50 per million tokens, 1M context window, 65K max output, 304 tokens/second, AA Intelligence Index score 50, knowledge cutoff March 2026. Google also launched Gemini 3.5 Flash-Lite ($0.30/$2.50/M) and a restricted security model called Gemini 3.5 Flash Cyber. Gemini 3.5 Pro remains delayed with no confirmed launch date.

Large Language Models22 Jul 20266 min read1,954 views
Top AI Tools:
ChatGPT Claude.ai Grok (xAI) Gemini DeepSeek Perplexity Midjourney GitHub Copilot Cursor ElevenLabs Runway Notion AI