Live Feed
MiniMax H3 Ships Open Weights: 33B Omni Video Model Challenges Closed API Leaders at $0.08/s
MiniMax releases H3 as open weights — a 33B diffusion transformer that accepts text, image, video, and audio inputs and outputs 4-15 second videos with native stereo audio. On Artificial Analysis Arena, it scores 1193 on I2V and 1237 on T2V, entering the leaderboard's top competitive tier.
Ramp July: Anthropic Reaches 43.5% of US Businesses as Fable 5 Stalls at 6% of Internal Tokens
Anthropic widened its business AI lead over OpenAI to 43.5% vs 39.7% in Ramp's July index — growing five times faster. But inside Anthropic's own token mix, Fable 5 accounts for just 6% of consumption one month after launch, and that share is not rising. Price is the constraint.
L&T Deploys India's First 10,000-GPU Single NVLink Cluster for Together AI in Chennai
Larsen & Toubro has secured a $1.05B-$1.57B contract to deploy 10,000 NVIDIA B300 Blackwell Ultra GPUs as one interconnected NVLink fabric at Vyoma.AI's Chennai campus — India's first unified AI supercluster, not a distributed collection of servers.
Google Takes Control of Crusoe's 1.8 GW Wyoming Campus — Project Jade Is Now Project Tembo
County records confirm Google has replaced Crusoe Energy as the development partner on a 716-acre Cheyenne campus with 42 natural gas turbines, fuel cells, and a $7B+ power investment targeting early 2028 service.
Cerebras Runs GPT-5.6 Sol at 750 Tokens Per Second — 7x Faster Than Fable 5, No Quality Loss
OpenAI and Cerebras launch Ultrafast Mode, a new API tier powered by Cerebras wafer-scale silicon that delivers GPT-5.6 Sol at 750 output tokens per second, completing Humanity's Last Exam in 11 hours versus Fable 5's 78.
DeepSeek Harness v0.1: MIT-Licensed Coding Agent With Swappable Plugins Enters the Claude Code Fight
DeepSeek open-sources Harness, a Codex and Claude Code rival built on a swappable plugin system called Cordis, as the lab simultaneously raises API cache prices 6x for agent-heavy workflows.
Gemini 3.7 Flash: DeepSWE Jumps 16 Points to 65.3%, Priced at Half of What 3.6 Flash Cost at Launch
Google ships Gemini 3.7 Flash three weeks after 3.6, with FrontierCode up to 43.6%, DeepSWE at 65.3%, and an introductory price of $0.75/M input — 50% cheaper than 3.6 Flash was at launch.
OpenAI Enterprise: 64% of Output Tokens Now Run Agent Loops — Chat Has Become the Minority Workload
New data from OpenAI shows that as of June 2026, agentic AI — defined as Codex tokens — accounts for 64% of combined enterprise output. The same frontier models power both chat and agents; the difference is how the compute gets spent.
White House Ties $54B Pentagon AI Budget to 30-Day Safety Tests — Open-Weight Models Next
White House met with Google, OpenAI, Anthropic, and Meta on August 12 to finalize voluntary AI safety framework. Labs must submit models for 30-day federal inspection to access Pentagon's $54B FY2027 AI allocations. Open-weight models expected to enter the framework once they reach Mythos-class capability levels.
Upstage Solar Pro 4 Jumps 28 Points on AA Intelligence Index — Terminal-Bench Up 4.8x, $0.30/$1.20 Per Million
Korean AI lab Upstage releases Solar Pro 4, scoring 42 on the Artificial Analysis Intelligence Index — up 28 points from Solar Pro 3's 14. Terminal-Bench v2.1 improves 4.8x to 57, τ³-Banking multi-turn tool use rises 2.5x to 23. Priced at $0.30/$1.20 per million input/output tokens.
OpenRouter's Live Search Benchmarks: Budget Beats Engine, Model Beats Engine, Failure Rate Drives Cost
OpenRouter published live leaderboards for web search agent configurations — engine times model times depth — and found search budget is the dominant quality variable while model choice outweighs engine selection in every combination tested.
Cognition Eyes $40B in New Talks — Devin Revenue Has Nearly Doubled Since Its $26B Close Three Months Ago
Bloomberg reports Cognition is in early discussions for a round valuing the company at $40 billion or more, up 54% from its May close. Annualized revenue is approaching $1 billion, against $492 million in May.
DeepSeek V4 Pro 0813 Goes GA After Four-Month Preview — 1.6T MoE Becomes the Official Flagship
DeepSeek ends a four-month preview for V4 Pro, designating the 0813 build as its production API release. The 1.6T-parameter MoE inherits April pricing: $0.435/M input, $0.87/M output, 1M context, 500 concurrent calls.
Grok 4.6 Reaches Intelligence Index 61 — SpaceXAI Matches GPT-5.6 Sol at $2 Per Million
SpaceXAI releases Grok 4.6, hitting AA Intelligence Index 61 to match GPT-5.6 Sol at a third of the cost. Built on the V9 1.5T foundation with improved RL and longer training, it leads GDPval-AA v2 at 1753 Elo among non-Anthropic models.
Lovable Raises $400M at $13.3B — Vibe-Coding Platform Doubles Its Valuation in Eight Months
Stockholm-based Lovable closes a $400M Series C co-led by Menlo Ventures and EQT, doubling its December 2025 valuation to $13.3B. The platform has 60 million projects created and 900 million monthly app visits from users at two-thirds of the Fortune 500.