Agent collaboration application Raft will open-source its code, as its founder admitted the startup deviated from its original path.

Beating AI Express News: Artificial Analysis has completed third-party testing of several new AI models released overnight. Claude Opus 5.5 max scored 58 points on the Intelligence Index v4.3, currently ranking first. GPT-6 Astra and Fable 5.1 both scored 53 points, while Opus 5 scored 51 points. Opus 5.5 max took the top score in 6 of the 10 tests. However, achieving these top scores required more tokens: Opus 5.5 max generates an average of around 119,000 tokens per task, roughly 60% more than Opus 5’s 73,000 tokens per task. Thanks to a 20% API price cut and 60% reduction in cache read costs, its per-task cost came to $5.98, almost on par with Opus 5’s $5.86. The more practical medium-tier model scored 51 points, matching Opus 5 max, with a per-task cost of just $1.34. OpenAI took a different approach: GPT-6 Sol and Luna delivered Intelligence Index performance close to their GPT-5.6 predecessor, but at significantly lower prices. Sol max’s per-task cost dropped from $1.99 to $1.06, while Luna’s fell from $0.18 to $0.07. In coding agent tests, Sol’s score rose from 55 to 57 points, with its cost cut by roughly half; Luna’s score dipped from 43 to 41 points, but its per-task cost is about 60% lower. The two companies’ strategies are starting to diverge: Opus 5.5 pushes performance limits higher with more inference, while GPT-6 Sol and Luna focus on making existing capabilities cheaper. On Artificial Analysis’s performance-cost chart, both companies’ new models occupy the new Pareto frontier.

Per Bitget market data, Hong Kong-listed AI application stocks extended their afternoon decline: Zhipu fell over 10%, Kingdee International dropped more than 7%, while Meitu, MINIMAX-W and other related stocks also tracked lower.

Whales keep buying $UNI today! A newly created wallet, 0xEFC4, bought 269,477 $UNI ($2.84M). 0xf415 bought another 138,442 $UNI ($1.23M).

Whale 0xA799 bought 788,000 $UNI ($4.93M) at $6.26 6 days ago and sold at $8.85, making $2.04M in less than a week!

Solana has begun testing the Alpenglow upgrade on its public testnet, with the goal of cutting transaction finality from roughly 12.8 seconds to 150 milliseconds. The upgrade will replace the TowerBFT consensus protocol with the Votor voting protocol, allowing validators to confirm blocks via one to two rounds of direct voting. Testing is currently only supported by the Agave 4.3 client developed by Anza; Jump Crypto’s Firedancer and Frankendancer clients have not yet been integrated. The mainnet launch date remains undetermined. (CoinDesk)

According to official announcements, NEAR has been deployed on Hyperliquid's spot market. Users can trade the NEAR/USDC pair on Hyperliquid. The official added that per Hyperliquid's procedures, NEAR will take several more days to appear on the platform's Strict List.