Liquid AI2026-09-25 23:19:21Liquid AI releases experimental draft model for faster vision-language decodingLiquid AI has introduced LFM2.5-VL-3B-DSpark, an experimental speculative decoding draft model built for its LFM2.5-VL-3B vision-language model. The company said the draft model adds about 280 million parameters and is designed to speed up decoding without changing model outputs. According to the Liquid AI team, decoding throughput improved by as much as 3.13x on Apple silicon and up to 2.66x on NVIDIA H100 hardware. The model has already been deployed, and its weights are available on Hugging Face in both Safetensors and GGUF formats. It also received day-one support across SGLang, MLX-VLM, and llama.cpp. Liquid AI labeled the release as experimental and said it is distributed under the LFM Open License v1.0, which allows free commercial use for companies with annual revenue below $10 million. The update was cited by Techub, referencing MarkTechPost.200
zero-knowledg2026-08-12 02:29:16Attestable Raises $20M Seed Round, Earns Vitalik Praise for ZK BreakthroughAI integrity company Attestable has announced a $20 million seed round and formally launched. Jamin Ball of Altimeter Capital and Yonatan Mandelbaum of TLV Partners co-led the round, with Halcyon Futures, Cerca Partners, Zero Knowledge Ventures and several individual investors participating. Ethereum co-founder Vitalik Buterin retweeted CEO Yogi's post, calling the project "impressive work." Attestable said it solved what was previously considered infeasible: practical zero-knowledge proofs. The approach shifts trust from the data center to a small mathematical verifier, allowing a data center to prove that approved models, weights, inputs and policies produced the corresponding output without exposing model weights or private data. Verification needs no private key and no rerunning of the model. On a single NVIDIA H100, the team's alpha version verifies Meta Muse Glimmer 30B at 85 tokens per second. The proof is short and post-quantum secure.2101
Aethir2026-07-15 23:29:43Aethir tokenizes GPU compute on InjectiveDecentralized GPU platform Aethir has tokenized the computing capacity of more than 43,000 GPUs on the Injective blockchain, including over 3,000 NVIDIA H100 units, according to Techub News. The move brings a large pool of distributed compute resources onto on-chain infrastructure built for financial activity. Aethir describes itself as a decentralized GPU computing platform focused on delivering enterprise-grade graphics processing resources through a distributed network. Injective is a Layer 1 blockchain built on the Cosmos SDK and geared toward DeFi applications. With the integration, users can trade and stake computing power in tokenized form. The arrangement links traditional compute resources with blockchain-based financial rails, extending the use of tokenization beyond digital-native assets into infrastructure tied to AI and high-performance computing.1730
