Lookonchain APP

App Store

DeepSWE Swaps Top Spot Twice Overnight: Gemini Just Claimed First Place, Meta Notches 75.4%

1 hours ago

Beating AI Express: Right after Gemini 3.8 Flash launched, Google scored 73.7% on DeepSWE v1.1. The DeepSWE official leaderboard now lists the model at 74%±1%, placing it ahead of Claude Opus 5 and GPT-5.6 Sol. Just hours later, Meta unveiled Muse Spark 1.3, which notched a 75.4% score on the same DeepSWE benchmark, pushing the public benchmark score up by an additional 1.7 percentage points. The test tasks agents to independently complete long-cycle software engineering tasks from 113 real codebases. Both Google and Meta used the mini-swe-agent framework: Google ran the high inference profile, while Meta used the max profile. Meta's previous-generation Spark 1.2 only scored 55.0%, marking a 20.4 percentage point jump this time around. However, Meta's 75.4% score has not yet been added to the official DeepSWE leaderboard. On Artificial Analysis's Coding Agent Index, the limited-preview Spark 1.3 max scored 68 points, second only to Claude Opus 5; the currently publicly available xhigh version of the model scored 64 points.

Relevant content

Ukrainian police have dismantled a cryptocurrency fraud ring in Kyiv, with the case involving up to $1 million in monthly illicit proceeds.

Ukraine’s National Police and Security Service have seized a fake cryptocurrency investment platform network based in Kyiv. The scam group used Telegram to distribute fake investment ads, luring victims to a counterfeit trading platform to steal their wallet assets. Investigations confirmed the group defrauded 62 victims across more than 20 countries, including Germany, Poland, France, the UK, Canada, Israel, and other regions. Fraudsters displayed false returns via forged trading interfaces; when users applied for withdrawals, they tricked them into authorizing small test transactions under the pretext of “account verification”, then exploited built-in crypto-theft programs to transfer funds from victims’ wallets. Ukrainian security authorities stated the criminal ring is led by a 25-year-old IT professional, with a peak monthly operation scale of $1 million. Police conducted 34 searches in Kyiv and its surrounding areas, seizing over 100 computers, more than 100 mobile phones, 79 SIM cards, and a large number of related devices. The case remains under further investigation, as police pursue additional suspects, identify more victims, and trace the full scale of stolen funds.

9 minutes ago

OpenAI Accelerates Embodied Intelligence Push, Sam Altman Says It 'Will Definitely Develop Humanoid Robots'

Beating AI News Flash: OpenAI CEO Sam Altman stated that the company "will definitely develop humanoid robots, and will also explore other forms of robots." Altman believes that since most real-world facilities and tools are designed around humans, humanoid structures are better suited to operate in physical environments. OpenAI is currently restructuring its robotics team, recruiting talents in areas including robot control algorithms, actuator design, data collection, and Embodied AI. Earlier around 2021, OpenAI shut down its early robotics project due to insufficient real-world data to train robot systems. Now, with advancements in large models and robot data infrastructure, the company is re-investing heavily in this field. Altman has repeatedly expressed interest in Embodied AI and humanoid robots before, noting that it would be a limitation if AI had near-general intelligence but was unable to perform tasks in the real world.

9 minutes ago

Bloomberg: Low volatility in US equities may signal risks, while gold's advantage over US Treasuries is near a historic high.

Bloomberg commodities strategist Mike McGlone wrote in a note that U.S. stock market volatility relative to gold is at its lowest level since 2007, and as markets enter their traditional volatile season, this situation could impact the performance of gold, stocks, and bonds in the second half of this year. McGlone noted that the ratio of the SPDR Gold ETF (GLD) to the iShares 20+ Year U.S. Treasury Bond ETF (TLT), which he tracks, is near an all-time high, indicating gold is performing extremely strongly relative to long-term U.S. Treasuries. He said that historically, extremely low stock market volatility occurred ahead of the 2008 financial crisis, and whether the current market will repeat a similar scenario remains to be seen. After gold surged to around $5,600 per ounce in the first quarter of this year, it may face pullback pressure similar to that seen after crude oil prices peaked in 2008. McGlone pointed out that commodity markets have a reversal effect after "rising too fast". After crude oil hit its peak in 2008, it weakened continuously relative to its 60-month moving average, with successive lower highs and lower lows. At that time, crude oil's premium relative to its long-term moving average hit its highest level since the 1973-1974 oil crisis. In February this year, gold once reached a premium of about 2.2 times its 60-month moving average, a level last seen in 1980. However, the difference is that this round of gold's rise has set an unprecedented record amid a non-high-inflation environment, so its subsequent trend remains to be watched.

9 minutes ago

International oil prices continue to climb, with both WTI and Brent crude up over 1%.

According to Bitget market data, both US and Brent crude oil prices rose by over 1%. WTI crude oil is currently trading at $90.01 per barrel, while Brent crude oil stands at $95.26 per barrel.

9 minutes ago

Claude E-commerce Agent open-sourced: Partner merchants see a 35% increase in shopping cart volume and a 60% rise in customer purchase rate.

Beating AI News Flash: Anthropic has open-sourced Claude Commerce Agents, a set of reference code enabling merchants to build their own shopping and operations agents. The suite includes two agents: one for consumers to search for products, compare items, bundle multiple goods, and add to cart; the other for merchants to monitor sales, inventory, pricing, and marketing. The code is licensed under Apache 2.0. Notably, the shopping agent only passes the cart to the merchant’s own checkout page—no direct payment processing interface is included. For the merchant agent, any adjustments to pricing, restocking, or promotions must first generate pending changes that require manual approval before implementation. These restrictions are not limited to prompts: rules around payments, product sourcing, pricing adjustment ranges, and manual approvals are enforced at the code level. Anthropic does not recommend splitting capabilities like search, returns, and pricing into multiple sub-agents. Instead, it uses a single Claude model that retains full conversation context, loading different skills on an as-needed basis. The company states that in comparisons across multiple enterprise deployments, this approach delivers higher quality, typically uses fewer tokens, and has lower latency. According to Anthropic, one partner saw a ~30%–35% increase in shopping cart size and a ~60% rise in customer purchase completion rates after deployment.

9 minutes ago

Robinhood Chain’s data hits a new record, with daily on-chain revenue surpassing $4 million, topping the public blockchain sector.

According to DeFiLlama data, Robinhood Chain’s 24-hour DEX trading volume hit approximately $1.851 billion, marking the sixth consecutive day it has set a new all-time high. Currently, this volume ranks second among all blockchains, trailing only Solana (around $2.531 billion) and outpacing Ethereum (around $1.32 billion), BSC (around $1.181 billion), and Base (around $800 million). Over the same period, Robinhood Chain’s on-chain fees stood at roughly $4.45 million, while its on-chain revenue reached about $4.01 million—both ranking first in DeFiLlama’s chain metrics. Breaking down the figures: Robinhood Chain’s 24-hour Chain Fees totaled around $4.45 million, while the combined fees of Solana, BSC, Ethereum, and Base came to roughly $1.49 million. In terms of revenue, Robinhood Chain’s 24-hour network layer revenue hit approximately $4.01 million, compared to just $288,000 in combined revenue from Base, Solana, Ethereum, and BSC. This means Robinhood Chain’s single-chain daily revenue is roughly 13.9 times the combined total of these four chains. Furthermore, the surging trading activity on Robinhood Chain has caught the attention of wallet projects. OKX Wallet announced today that users trading Robinhood Chain tokens via OKX’s built-in DEX will receive a limited-time full gas fee subsidy. Binance Wallet also announced the launch of a 20% fee reduction promotion for Robinhood Chain today.

9 minutes ago

Popular tokens

BitcoinEthereumHyperliquidSolanaTRONBNBTetherAaveXRPPepeFartcoinOndoJupiterUniswapBonkPendleEthenaArbitrumAvalancheLidoChainlinkPolygonDogecoinCardano