GPT-56T 861 —
MUSE-SPK 835 -0.7%
GPT-56SC 828 -5.2%
QWEN-38X 824 —
CL-OP55X 822 —
GROK-46H 822 -5%
GPT-6A 820 —
GLM-5 784 -8.4%
CL-FAB5H 743 -5.6%
KIMI-K3X 742 -8.4%
CL-OP5H 720 -5.8%
CL-OP5X 709 -18%
CL-OP46H 698 -5.9%
CL-OP47H 690 -5.9%
GEM-38FH 677 +0.1%
GEM-37FH 657 -24%
GPT-56S 622 —
CL-OP47 582 -0.7%
GPT-55H 582 —
INKL 531 —
GEM-31P 513 —
GEM-3P 499 —
CL-OP46 496 -0.2%
CL-OP48 490 —
GPT-56T 861 —
MUSE-SPK 835 -0.7%
GPT-56SC 828 -5.2%
QWEN-38X 824 —
CL-OP55X 822 —
GROK-46H 822 -5%
GPT-6A 820 —
GLM-5 784 -8.4%
CL-FAB5H 743 -5.6%
KIMI-K3X 742 -8.4%
CL-OP5H 720 -5.8%
CL-OP5X 709 -18%
CL-OP46H 698 -5.9%
CL-OP47H 690 -5.9%
GEM-38FH 677 +0.1%
GEM-37FH 657 -24%
GPT-56S 622 —
CL-OP47 582 -0.7%
GPT-55H 582 —
INKL 531 —
GEM-31P 513 —
GEM-3P 499 —
CL-OP46 496 -0.2%
CL-OP48 490 —

Live Feed

3mo ago research

Shift Offers Free NYC Apartment Cleanings in Exchange for Robot Training Data — 10,000 Operators, 15 Countries

MicroAGI's Shift app sends hat-cam operators into homes to collect first-person embodied AI training footage, then licenses it to robotics companies. The free cleaning is the barter. The company paid 10,000+ people $5M in Q1 2026.

3mo ago release

Robinhood Becomes First Major Broker to Let AI Agents Trade and Shop — Dedicated MCP Accounts, 3% Cash Back

Robinhood launched Agentic Trading and an Agentic Credit Card on May 27, opening both brokerage and spending rails to third-party AI agents via MCP. Stock trading is live in beta; options, crypto, and futures follow.

3mo ago policy

UK to Deploy AI Facial Age Estimation at Borders From 2027 — 43% of Child Claimants Were Adults

The UK Home Office has awarded a contract to build and test AI facial recognition for age-estimating asylum seekers, citing data showing 43% of migrants claiming to be children were found to be adults after manual assessment.

3mo ago policy

Mistral CEO Rejects Pope Leo's AI Disarmament Call, Announces €4B Infrastructure Plan

Three days after the Vatican published Magnifica Humanitas calling for AI to be 'disarmed' from warfare, Mistral's Arthur Mensch said Europe cannot afford to stand down and announced a €4B infrastructure push targeting 1GW by 2030.

3mo ago release

Liquid AI's LFM2.5 Triples Training Data to 38T Tokens and Cuts Hallucinations by 9x on Device

Liquid AI's new on-device MoE model posts 88.76% on MATH500 and 42.53% on AIME25 with 1B active parameters, while pushing its non-hallucination rate from 7% to 63% through targeted RL.

3mo ago research

Image Diffusion Trainers Overlooked the Residual Stream for Years — New Paper Fixes It in 8.75x Fewer Iterations

A paper on arxiv (2605.20708) introduces Diffusion-Adaptive Routing, replacing the static residual stack in diffusion transformers with per-timestep layer routing. Result: the same image quality in 8.75 times fewer training iterations. The fix targets a bottleneck everyone walked past.

3mo ago policy

Nvidia Can't Sell Chips to China — So Jensen Huang Joined Beijing's Top Business Advisory Board Instead

Huang accepts a seat on Tsinghua University's SEM advisory board, chaired by Tim Cook, alongside Musk, Nadella, Zuckerberg, Dimon, and Fink. The timing: days after Trump's China state visit, while Nvidia's China data center revenue sits at zero.

3mo ago research

AgingBench: Deployed AI Agents Degrade Without Model Updates — UT Austin Paper Names 4 Ways Memory Rots

A University of Texas study finds AI agents become measurably less reliable after deployment, even when the underlying model is unchanged. The paper introduces AgingBench, a cross-session reliability evaluation, and identifies four distinct failure modes that compound over time: summary drift, memory blur, stale facts, and maintenance-induced breaks.

3mo ago research

Global Humanoid Shipments Rose 800% in 2025: China Fields 140 Manufacturers as AGIBOT Claims 39% Market Share

IDC data shows global humanoid robot shipments grew nearly 800% in 2025. China now has 140 humanoid manufacturers with 330 new models released in 12 months. AGIBOT shipped 5,200 units last year — 39% of global volume — as the first mass-delivery lines for competing models go live.

3mo ago benchmark

Claude Opus 4.8 Leads Intelligence Index at 61.4: AA Analysis Puts GDPval Elo at 1890, 67% Win Rate Over GPT-5.5

Artificial Analysis independent benchmarking places Claude Opus 4.8 at 61.4 on the Intelligence Index, +1.2 ahead of GPT-5.5 xhigh. On GDPval-AA, the primary agentic evaluation, Opus 4.8 scores 1890 Elo — implying a 67% win rate against GPT-5.5 while using 35% fewer output tokens. Pricing is $5/$25 per million tokens, the same as Opus 4.7.

3mo ago funding

CME Group and DRW Launch GPU Compute Futures — Shanghai Designs AI Token Derivatives in Parallel

CME Group and Silicon Data (backed by DRW) will launch the first compute futures contracts later this year, using real-time GPU price indices as the underlying. Intercontinental Exchange is building a rival product. China's Shanghai Futures Exchange is taking a different path: derivatives tied to AI inference tokens, not GPU rental rates.

3mo ago research

Tencent's Hy3 Is Beating Claude by 50% on OpenRouter — At $0.066/M, Nobody Can Explain It

Hy3 preview, a mid-tier Chinese open-weights MoE model, is processing 3.47 trillion tokens weekly on OpenRouter, outpacing Claude Opus 4.7 by more than 50%. It is not the cheapest model available. Its benchmarks are not frontier-tier. The data does not offer a clean answer.

3mo ago release

Claude Code Dynamic Workflows Run Hundreds of Parallel Subagents: Quarter-Length Work Now Takes Days

Anthropic launched Dynamic Workflows in Claude Code, enabling a single session to orchestrate tens to hundreds of parallel subagents. Available in research preview on Max, Team, and Enterprise plans, and on the API, Bedrock, Vertex AI, and Microsoft Foundry. Activates via ultracode mode or explicit workflow command.

3mo ago model

Claude Opus 4.8 Launches: Super-Agent Benchmark Leader, 84% on Mind2Web, Fast Mode 3x Cheaper

Anthropic released Claude Opus 4.8, the only model to complete every case end-to-end on the Super-Agent benchmark — beating GPT-5.5. Browser agent performance hits 84% on Online-Mind2Web. Fast Mode runs at 2.5x speed and is now three times cheaper than previous Opus fast pricing. Available today at Opus 4.7 pricing.

3mo ago funding

Anthropic Raises $65B in Series H at $965B, Overtakes OpenAI as Most Valuable AI Lab

Anthropic closed a $65 billion Series H led by Altimeter, Sequoia, Dragoneer, and Greenoaks, pushing post-money valuation to $965 billion — above OpenAI's $852 billion. Run-rate revenue crossed $47 billion this month. Micron, Samsung, and SK hynix join as strategic memory partners.