1inch联合创始人Anton Bukov宣布已离开项目
Related
OpenRouter introduces the Fusion composite model: combining multiple models in parallel, offering a price comparable to Fable 5 at half the cost.
According to Beating's monitoring, OpenRouter has launched the Fusion composite model interface, which supports the parallel distribution of user prompts to multiple large language models, and then outputs the final solution through a referee model and a synthesis model. In the DRACO deep learning benchmark released by Perplexity AI, multiple model combinations based on Fusion demonstrated their superior performance compared to traditional single models. In various configurations, combining Fable 5 and GPT-5.5 into a panel and synthesizing it with Opus 4.8 achieved the highest score of 69.0%, significantly surpassing the 65.3% score of Fable 5 alone. The test shows that due to the differences in the underlying training and logic of models from different vendors, the diversity of multi-model combinations can bring stronger complementary perspectives in complex tasks. Even the same model performing self-combination showed a significant increase; combining Opus 4.8 with itself into a dual panel and synthesizing it increased the score from 58.8% to 65.5% when running independently. For high cost-effectiveness requirements, a low-cost panel consisting of Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro also achieved a score of 64.7% in synthesis, narrowing the gap with Fable 5 to 0.6 percentage points while halving the call cost. The benchmark test includes 100 complex research tasks across 10 dimensions. To prevent the model from accidentally accessing online evaluation standards during retrieval, which could distort the score, OpenRouter has configured a list of filtered search domains on the server side. Fusion is now open to API users, with the default identifier being openrouter/fusion. Users can also customize the model panels used in synthesis on the web interface.
TianTong Co., Ltd.: 6-inch and 8-inch lithium niobate wafer materials have entered mass production, and 12-inch wafers have completed R&D verification.
According to Mars Finance, TianTong Co., Ltd. (600330.SH) released an investor relations activity record announcement, stating that the company's 6-inch and 8-inch lithium niobate wafer materials have been mass-produced, and the 12-inch wafers have also completed R&D verification. Lithium niobate crystal materials are expected to benefit from the technological transformation in the optical communication industry during the "fiber-to-the-home" transition. According to industry research, the technology route of using lithium niobate materials in the modulation and conditioning circuits of high-speed optical modules of 3.2T and above is expected to become the mainstream solution for downstream customers. The company's integrated inductors are currently in the early stage of large-scale growth, and the product technology has been recognized by leading customers. Production lines are gradually being implemented, and with the continuous increase in the power consumption of data center chips, market demand is showing an upward trend. (Cailian Press)
The Goldfinch community proposal to enter maintenance mode and gradually liquidate Prime services currently has a 100% approval rating.
According to Odaily Odaily, the decentralized lending protocol Goldfinch community has released a proposal to "Approve the operation and maintenance of Goldfinch and the gradual liquidation of Goldfinch Prime." Voting will end tomorrow, and the current approval rate is 100%. The proposal includes the orderly closure of Goldfinch Prime operations and the transition of Goldfinch to "maintenance mode." Subsequent priorities will include supporting the continued liquidation of historical borrower pools, maintaining access to previous versions of the application so users can continue to receive repayments and principal, and fulfilling the foundation's closure and related administrative obligations. In addition, Warbler Labs will receive a fixed service fee of $150,000, of which $100,000 is new budget and $50,000 comes from the reallocation of previously approved operating budget. These funds will be used to support the liquidation and business transition of Goldfinch Prime and the Foundation, as well as the technical and operational maintenance of the legacy loan pool over the next two years or longer. Previously, Goldfinch completed a $25 million funding round in 2022, led by a16z.
Canton Strategic launches $50 million stock buyback program
Odaily Odaily reports that Nasdaq-listed Canton Coin treasury firm Canton Strategic announced its board of directors has approved a $50 million stock buyback program. The buyback will reportedly be conducted in accordance with U.S. securities laws through the open market or other compliant means. However, the timing, size, and method of the buyback may be adjusted or terminated based on market conditions, stock price performance, trading volume, and regulatory environment, and no minimum or fixed buyback amount is guaranteed. (PRNewswire)
Bitwise launches trading product "BWCC" on the Canton exchange.
Odaily Odaily reports that Bitwise announced the launch of Bitwise Canton ETP (BWCC) by its subsidiary Bitwise Europe GmbH, and its listing on Deutsche Börse Xetra, an electronic spot trading platform under the Deutsche Börse. This allows trading through regular brokerage accounts, aiming to provide trading exposure to the Canton Network's native token, CC. (Globenewswire)
Gate Prediction Market: Antonelli and Russell Lead Canadian Grand Prix Winner Prediction
According to the latest data from Odaily, in the "Canadian Grand Prix: Drivers' Championship" market, Kimi Antonelli currently leads with a 34% probability, followed closely by George Russell at 32%. Lando Norris has 14%, while Max Verstappen and Oscar Piastri are both at 9%. As the first CEX platform to integrate with Polymarket, Gate has consistently ranked among the top three in Polymarket's partner channels, with significant growth in user participation, reflecting the increasing activity and depth of user engagement in the prediction market. Currently, users can directly access the prediction market through the Gate App, entering the Polymarket page from the platform's homepage alpha and participating in event predictions using USDT from their exchange accounts. Gate continues to optimize its prediction market functionality and ecosystem, providing users with a diversified market experience from prediction and trading participation.