24-hour Global Liquidations Soar to $1.02 Billion, Bitcoin Hits Lowest Price Since Early July

According to YuEmber Monitoring, a whale or institutional entity accumulated 37,000 ETH via Galaxy Digital two months ago at an average price of $1,923. Today, it added another 15,000 ETH at an average price of $2,751, worth approximately $41.26 million. Its current total holdings of 52,000 ETH (valued at roughly $143 million) have an average cost basis of $2,161, generating an unrealized profit of $31.1 million.

According to Yu Jing Monitoring, the asset liquidation team of FTX/Alameda Research took advantage of the recent crypto market rebound to transfer 27,372 ETH (worth approximately $75.32 million) to Wintermute via six wallets three hours ago, likely for consigned sale.

According to Coinglass data, ZEC has broken through $1,600, notching a new recent all-time high. In the past four hours, total liquidations across the network reached $13.4047 million, ranking first, with short-position liquidations accounting for $12.9 million.

According to monitoring by Onchain Lens, the DEED token on Robinhood Chain is suspected of a rug pull. The token’s issuance is highly concentrated: at its peak, 110 associated wallets controlled approximately 86% of the total supply and have collectively withdrawn around $700,000 in funds. The project’s wallet also claimed 68.5 ETH (valued at about $188,900) in creator fees. DEED’s market capitalization has plummeted from a peak of $4.23 million to roughly $55,000, a decline of 98.7%.

CryptoQuant founder Ki Young Ju stated in a post that he expects Bitcoin’s current bull run to surge 3 to 5 times, rather than another parabolic rally of over 10x, followed by a more moderate bear market. As Bitcoin’s market capitalization expands and institutional holdings rise, the extreme rallies once dominated by retail investors and hot money, along with 80%+ crashes, are being suppressed. Factors limiting upside are also mitigating downside risks. The PNL index shows that the extremes of Bitcoin’s cycle tops and bottoms are decreasing, while overall position profitability during bottom formation continues to rise. In this cycle, the MVRV ratio never fell below 1; even at market lows, Bitcoin’s price remained above holders’ average on-chain cost basis. Moreover, realized market capitalization keeps climbing, signaling new capital inflows. Early whales have ceased selling, and futures market whales have built substantial long positions near the bottom. The 365-day moving average of the PNL index is also forming a clear inflection point. This does not mean Bitcoin has an upside cap, but rather that the market’s risk-reward structure is evolving. Giving up a 10x+ parabolic rally also means avoiding an 80% extreme drop—this lower volatility could attract more long-term-focused capital instead of short-term hot money. He added that Bitcoin’s development path may gradually approach Satoshi Nakamoto’s vision: becoming a stable enough asset to truly function as a currency. As Bitcoin matures further, internet-native capital could reshape the global economy in ways exceeding expectations.

Beating AI News: OpenAI has launched GPT-6 Sol and GPT-6 Luna, extending GPT-6 capabilities to two lower-cost model tiers. Sol is designed for complex coding and professional tasks, while Luna targets high-frequency daily use cases. API pricing per million tokens: Sol costs $2 for input and $10 for output; Luna is priced at $0.10 per input token and $0.50 per output token. This marks at least a 50% discount compared to GPT-5.6’s promotional pricing. On the same day, Claude Opus 5.5 was priced at $4 input / $20 output, meaning Sol’s base token price is exactly half that of Opus 5.5. In OpenAI’s AutomationBench, Sol scores 33.2% at xhigh effort, outperforming Opus 5’s 26.9%, with a per-task cost of just $0.27—roughly 9% of Opus 5’s per-task cost. Note: this comparison is to Opus 5, not the newly released Opus 5.5. On DeepSWE v1.1, Sol scores 68.8% at max effort, trailing Fable 5’s top score by only 1.1 percentage points while boasting ~80% lower per-task cost. Security benchmarks also show significant improvements: in tests designed to induce models to falsely report coding work, Sol’s deception rate fell from 10.4% (for GPT-5.6 Sol) to 1.3%, Luna’s dropped from 9.5% to 2.8%, and GPT-6 Astra has a 0.5% deception rate. Both models are now available via API, ChatGPT Work, and Codex. Free and Go tier users can access Luna on desktop; it is not yet integrated into Chat. GPT-6 cache reads also come with a 90% discount, and adjusting reasoning effort or toggling tools will not invalidate existing caches.