NaiveAI releases open-source model with 1M token context support

iconKuCoinFlash
Share
AI summary iconSummary
NaiveAI, founded by Tsinghua University’s Dai Jifeng, has launched the open-source model Naive-N0.5-Flash under the MIT license. Built on Xiaomi’s MiMo-V2.5 Base, the model supports a 1 million token context length and features an enhanced attention mechanism. AI managed code, experiments, and optimization across 151 rounds over six days. This token launch underscores the model’s efficiency in processing long-form text, and new token listings on exchanges may follow as adoption increases.
ME AI News: NaiveAI, founded by Tsinghua University associate professor Dai Jifeng, has released Naive-N0.5-Flash and open-sourced its model weights. The model is based on Xiaomi’s MiMo-V2.5 Base and is open-sourced under the MIT license. NaiveAI retained the original MoE architecture but focused on modifying the attention mechanism. The team replaced global attention with sliding window attention and DeepSeek sparse attention, then continued training on 3.25 trillion tokens. The upgraded model natively supports up to 1 million token context lengths, significantly reducing computational load when processing extremely long texts. AI also directly participated in the development process—handling coding, running experiments, analyzing results, and further optimization—while researchers set the direction and made key decisions. For example, with the inference system NaiveRT, the team ran 151 optimization rounds in six days, 63 of which were adopted. The officially reported peak inference speed is 2122 tok/s, but this was achieved under specific conditions: using 8 GPUs, with thinking mode disabled, input processing time excluded, and measuring only the best one-second performance out of 41 requests. Under standard conditions, the official speed is approximately 50 tok/s per user. (Source: BlockBeats)
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.