v4

Coinbase tokenized stocks have launched on the Aave V4 Base deployment version. Eligible users from jurisdictions outside the United States can use AAPLc, AMZNc, GOOGLc, METAc, MSFTc, NVDAc, and TSLAc as collateral to borrow USDC.This V4 instance is deployed by Aave Labs on behalf of Aave DAO, with risk analysis completed by LlamaRisk. Chainlink serves as Coinbase's official oracle solution, providing price data for tokenized stocks that includes the underlying stock prices and issuer multipliers.

Aave V4 launched Equities Hub on Base, allowing eligible non-U.S. users to use seven U.S. stock tokens issued by Coinbase as collateral to borrow USDC, covering Apple, Amazon, Alphabet, Meta, Microsoft, Nvidia, and Tesla. The market is powered by Chainlink, providing on-chain price data for tokenized stocks. The initial collateral cap for the seven stocks is approximately $29 million, with a USDC supply cap of $32 million and a borrowing cap of $21 million. The collateralization rates for different stocks range from 65% to 79%.Currently, these stocks can only be used as collateral. Aave stated that more Coinbase tokenized stocks and GHO may be added as borrowable assets in the future after governance and risk assessments are approved.

The founder and CEO of the decentralized lending protocol Aave, Stani Kulechov, responded to various statements regarding Aave V4. He stated that the Hub and Spoke of Aave V4 are by default isolated according to risk characteristics, allowing risk-adjusted markets to share liquidity within set limits through the Hub to support new use cases while avoiding unnecessary fragmentation of liquidity. He believes that completely isolating liquidity often splits funds, reduces utilization, and increases user costs, which becomes more apparent after the incentives used to guide liquidity end.Stani mentioned that in V4, Spoke represents the lending market, and the Hub can share liquidity between these markets, allowing isolated risk characteristics to utilize the capital pool, thereby improving utilization and capital efficiency. Unlike typical curation vaults that start with zero liquidity and require funds or incentives to guide them, V4 Spoke can rely on the entire Hub balance sheet from the first day of launch. He also stated that the V4 architecture is simpler and remains flexible, with the overall codebase significantly smaller than Aave V3.According to him, Aave V4 has secured $1.2 billion in deposits and is deployed across multiple networks, including Ethereum, Avalanche, and Arc. V4 has supported third-party curators like EtherFi, allowing curators to build and manage complete market structures and participate in the economic distribution of a broader lending market, rather than just charging fees for managing deposit scales.

According to Token Terminal (@tokenterminal), the deposits of Aave v4 have exceeded 1 billion dollars, doubling in the past month, with the current active loan amount reaching 300 million dollars.

On September 14, Aave Labs released an ARFC on the Aave governance forum, seeking approval to deploy a new Aave V4 Isolated Hub and Spoke to integrate institutional custody collateral as collateral for stablecoin loans. Institutional borrowers will deposit collateral with Anchorage and maintain custody during the loan period, with the custody balance represented on-chain by non-transferable custody collateral tokens CoCT, minted and burned by CustodySync designed by Chainlink.Borrowers will stake CoCT on the Spoke and withdraw stablecoins from the Isolated Hub. The entire loan lifecycle, including liquidation, is synchronized between the custodian and Aave through Chainlink infrastructure. The proposal is limited to a single Isolated Hub and a single Spoke governed by the Aave DAO, without altering existing Hubs, Spokes, or reserves. The underlying assets will always remain in custody, with Anchorage holding the assets as the custodian and acting as the counterparty for account control protocols; Chainlink does not hold the assets.CoCT is a restricted transfer ERC-20, only allowing Hub, Spoke, and CustodySync, with minting and burning exclusive to CustodySync, and each borrower's position corresponding to a contract. The initial plan is to launch a CoCT for BTC held in custody by Anchorage. The next steps involve ARFC discussions, and if supported, it will proceed to Snapshot, followed by submitting the final parameters for AIP.

Decentralized Trading Infrastructure 0x stated that this year it has routed 81.92 million transactions with a trading volume of 4.267 billion USD, with approximately 70% touching Uniswap liquidity. 0x indicated that there has been an increase in malicious Uniswap v4 hooks recently, which provide one price in quote requests and another price at settlement, profiting from users by deceiving aggregators, wallets, and trading applications.After analyzing 84,163 hooks across 6 chains, 0x reported that only 19.4% are safe, 54.2% are malicious, and 26.4% may be malicious. Transactions routed through malicious v4 hooks can result in actual amounts received being up to 50% less than the quoted price. The article cites examples where the Base chain ETH/NVDAc pool hooks have accumulated fees of approximately 143,000 USD, and the BNB chain USDT/WBNB pool hooks have accumulated about 18,600 USD.0x stated that routing parties must ensure that the quoted amounts in pools are consistent with actual execution behavior, applications must have the capability to quickly eliminate suspicious routes, and users should be aware that optimal quotes are only useful when the route is safe. 0x has implemented measures such as detection and pool audits to prevent such pools from appearing in routing.

DeepSeek officially released the native multimodal MoE large model DeepSeek-V4.1-Flash. B.AI announced an additional 10% discount on the official new pricing from DeepSeek, allowing users to pay only 10% of the official price. This benefit will take effect on September 12 at 10:00 (SGT).B.AI will gradually route API requests from DeepSeek-V4-Flash and DeepSeek-V4-Flash-Vision-Exp to DeepSeek-V4.1-Flash, and users can also directly configure access to DeepSeek-V4.1-Flash.

The B.AI platform announces the official launch of the DeepSeek native multimodal MoE large model DeepSeek-V4.1-Flash. This model uses a backbone with 552 billion parameters and a Causal Encoder-Decoder architecture, with asymmetric computational scales for input and output.The model supports a context of 1 million tokens and 384K output, aimed at long-cycle programming agents, multimodal workflows, and high-concurrency agent systems. Currently, both Web Chat and API access have been opened.

According to the DeepSeek Harness team, the new version DeepSeek Harness v0.1.5 has been deeply debugged with the DeepSeek V4.1 Flash model, covering standard mode, Programmatic Tool Calling (PTC) mode, and minimalist mode, and supports updating system prompts while retaining the existing KV Cache.The web version adds support for uploading various file types such as images and PDFs, as well as a sidebar file tree preview, allowing direct viewing of Markdown, HTML, PDF, code, and images. The interface provides standardized extension entry points for plugins on both the left and right sides, supporting global entry, multiple tabs on the right, and column display.Long session loading and storage performance have been optimized, and bidirectional communication between the parent Agent and child Agents is more flexible, with the experimental plugin "Agent Teams" enabling multi-Agent team collaboration and task splitting.

DeepSeek has adjusted the offline schedule for V4 Pro. Previously, the official announcement stated that once V4.1 Flash officially goes live, all V4 Pro requests would be automatically switched to the new model; the latest notification changes this to September 14 at 12:00, when V4 Pro will be taken offline, and all requests will be switched to V4.1 Flash, with billing based on Flash prices.The previous automatic replacement plan has sparked opposition from many developers. Many developers are concerned that their carefully tuned Prompts, Agents, and production processes around V4 Pro will encounter issues due to the backend directly switching models. Some have suggested that DeepSeek retain the old model for a period of time, allowing users to decide when to migrate.

Less than 24 hours remain until the new pricing for the DeepSeek V4 Flash series takes effect (September 10, 12:00 SGT). With upstream price adjustments imminent, B.AI has prepared an exclusive stacking benefit for global developers to ensure top-tier computing power remains within reach. This upcoming benefit aims to further reduce invocation costs and unleash greater productivity—whether for high-frequency API calls, Agent workflow construction, or large-scale production-level deployments, all can operate continuously and efficiently with a lower burden. Specific benefits will be announced soon, so please stay tuned to B.AI's official updates and don't miss this boost in computing power.

DeepSeek V4.1 Flash has started launching the app, merging the original three modes: "Quick," "Expert," and "Image Recognition" into a single entry point. Users no longer need to manually select a mode; one model handles daily conversations, image understanding, and complex questions simultaneously.V4.1 Flash natively supports multimodal capabilities. DeepSeek previously stated that the new model has comprehensively surpassed V4 Pro in performance, cost, speed, and total time metrics, and will take over all V4 Pro requests before the launch of V4.1 Pro.

DeepSeek plans to officially release V4.1 Flash around September 10. The official statement claims that after internal and external testing, the new model has comprehensively surpassed V4 Pro in terms of performance, cost, speed, and total time.After the launch of V4.1 Flash, all requests sent to V4 Pro will automatically switch to V4.1 Flash before the release of V4.1 Pro, and will be billed at the cheaper Flash price. Users do not need to modify the original V4 Pro calling method.

DeepSeek has started internal testing of version 4.1 Flash intermediate version. The official WeChat group notification states that the new model has a new architecture, natively supports multimodal capabilities, is more powerful, faster, and cheaper. The current billing is the same as V4 Flash, with a maximum of 20 concurrent sessions per account.Users can now directly call it, and there is no need to change the base_url, just replace the original model ID with deepseek-v4.1-flash-expires-on-0910.

According to a post by X platform user "Yongchutafi" (@xhyctf) on September 8, DeepSeek may launch a new model V4.1 Flash.

On September 3, the B.AI platform DeepSeek-V4-Flash and DeepSeek-V4-Flash-Vision-Exp discount officially took effect. The platform adds a 50% discount on top of the DeepSeek official peak and valley pricing mechanism, with peak hours at 50% of the official price and off-peak hours as low as 25% of the official peak price.At the same time, the models GLM-5.3-Flash, Qwen3.8-Flash, Hy3, and MiMo V2.5 remain 100% free to access.

The AI model platform B.AI announced that starting from September 3, 2026, at 17:00 (SGT), the free trial of DeepSeek V4 Flash and DeepSeek V4 Flash Vision Exp will end, and a 50% billing will officially begin, while the previous two models will still be available for free.B.AI will synchronize the official peak and valley pricing of DeepSeek and implement additional discounts: during peak hours (weekdays 09:00--12:00, 14:00--18:00 SGT), billing will be at 50% of the official original price, and during off-peak hours (the remaining hours on weekdays and all day on weekends), an additional 50% discount will apply, making it 25% of the official original price. GLM 5.3 Flash will remain free, featuring a total of 320B parameters and 18B active parameters, supporting 1M context. Qwen3.8 Flash, Hy3, and MiMo V2.5 will also continue to be available for free.

The Aave on-chain dashboard shows that the current active loan scale of Aave V4 is $216 million. The V4 deployment covers Ethereum, Optimism, and Avalanche, with Ethereum Core deposits at $378 million, EtherFi Cash on Optimism at $257 million, Ethereum Global Dollar at $75 million, and Ethereum Prime at $63 million.User deposits for Aave V3 still reach $31 billion, with Ethereum Core at $25 billion. Aave's documentation states that V4 adopts a hub-and-spoke system to centralize liquidity and accounting, and sets independent borrowing rules and risk limits for different markets.

The Aave on-chain dashboard shows that V4 user deposits have reached $806 million, a 30% increase over 7 days, setting a new historical high. On August 19, deposits surpassed $500 million, on August 21 surpassed $600 million, and then within six days surpassed $800 million; on August 25, Ethereum V4 deposits alone exceeded $500 million.V4 active loans reached $206 million, with the EtherFi market having active loans of $62 million, and weETH collateral supporting WETH lending, with a utilization rate of 92%. The current market covers Ethereum, Optimism, and Avalanche deployments: Ethereum Core holds $378 million, EtherFi Cash on Optimism is $257 million, Ethereum Global Dollar is $75 million, Ethereum Prime is $63 million, Avalanche Core is $18 million, and Ethereum Plus is $15 million.In terms of deposit composition, weETH leads with $97 million, USDG $90 million, WETH and USDC each $81 million, liquidETH $77 million, liquidUSD $58 million, and WBTC $54 million. In comparison, V3 deposits reached $31 billion, while V4 replaces V3's market-per-pool design with a hub-and-spoke system.

Zhipu released the lightweight flagship model GLM-5.3 Flash, aimed at competing with DeepSeek V4 Flash, and utilizes over 100,000 domestic chip clusters to provide inference services. The model has a parameter scale of approximately 300B, about 40% of GLM-5.3, and is roughly on par with DeepSeek V4 Flash; it adopts the MoE architecture, with actual activated parameters around 18B, possesses native multimodal capabilities, supports image and video input, and is the first new model with multimodal capabilities following Zhipu's focus on coding strategy.The comprehensive evaluation score exceeds that of the previous flagship GLM-5.2, is on par with Claude Opus 4.8, and is higher than the relevant scores of DeepSeek V4 Pro official version. The pricing is 0.8 yuan per million tokens for input, 2.8 yuan for output, and 0.23 yuan for cache hits, approximately one-tenth of GLM-5.3, with a half-price discount offered for the first two weeks after launch. The model is fully compatible with domestic computing power, online traffic is handled by domestic chips, and Zhipu claims that the cluster hardware efficiency and unit token cost have reached levels comparable to mainstream NVIDIA GPUs; this model was previously tested anonymously on overseas platforms under the code name Ox Alpha.