AI·THE EXPLAINERニュースを理解する
StepFun previews Step 5 with open weights coming Oct 15Step 5 Preview公開、600B パラメータモデル
StepFun unveiled Step 5 Preview, a sparse Mixture-of-Experts model with 600B total parameters and 27B active per token, to be released with open weights on October 15.Stepfunが「Step 5 Preview」を発表。600Bのパラメータを持ち、1M トークンのコンテキストウィンドウとビジョン入力に対応し、10月15日にオープンウェイトで公開予定。
THE SOURCE TRAIL情報の根拠
1 first-party source linked. 0 news organizations linked.当事者・公的資料1件、報道機関0件にリンクしています。
Inspect the sources ↓出典を確かめる ↓What happened何が起きたか
StepFun announced Step 5 Preview, built on a sparse Mixture-of-Experts architecture with 600B total parameters and 27B active per token. The model supports a 1M-token context window and vision input. Step 5 Preview scores 44 on the Artificial Analysis Intelligence Index. StepFun said the model will be released with open weights on October 15.StepFunはStep 5 Previewを発表しました。必要な部分だけを動かすMixture-of-Experts構成で、総パラメーター数は6,000億、各トークンの処理で使うのは270億です。100万トークンの文脈と画像入力に対応します。Artificial Analysis Intelligence Indexでは44を記録しました。StepFunは、10月15日にモデルのウェイトを公開するとしています。
Why it matters何が変わるのか
The release adds another large-scale open-weight model to the field, giving developers and researchers a new option with a large context window and vision support. Hacker News commenters noted the model's active-parameter count and questioned its inference efficiency compared with other open models mentioned in the discussion, such as GLM 5.3 flash and DeepSeek V4/4.1 flash.開発者や研究者にとって、長い文脈と画像入力に対応する大規模なオープンウェイトモデルの選択肢が増えます。Hacker Newsのコメントでは、実際に動かすパラメーター数に注目が集まり、GLM 5.3 flashやDeepSeek V4/4.1 flashなど、議論で挙がったほかの公開モデルと比べた推論効率を疑問視する意見もありました。これはコメントでの反応です。
What’s nextこれからの動き
The open weights are scheduled to be released on October 15.ウェイトの公開は10月15日の予定です。
A little context · terms in this story理解の手がかり · この記事の用語
- Open weights
- The model’s learned parameters are available to download. The license still determines how you may use or redistribute them; open weights does not necessarily mean open source.AIが学習したパラメーターをダウンロードできること。利用・再配布の条件はライセンスで決まり、オープンソースと同じ意味とは限りません。
- Token
- A unit of text processed by a language model, often a word or part of a word. Input and output tokens may have different prices.言語モデルが文章を処理する単位。単語や単語の一部などに分かれ、入力と出力で料金が異なる場合があります。
- Inference
- Running a trained model to produce a result from new input, as distinct from training the model.学習済みのモデルに新しい入力を渡して結果を得る処理。モデルを学習させる作業とは区別します。
Read the sources原典を読む
These are the source links saved with this story. Older records do not identify which texts were used in the summary.この記事に保存されている出典です。過去の記事には、どの本文を要約に使用したかの記録がありません。
More in AIAIのほかの記事
SpaceXAI releases Grok 4.7 for coding and knowledge workSpaceXAI、コーディングと知識労働向けにGrok 4.7を公開
3 linked sites関連3サイト
Up to 10 references参照画像は最大10枚
Generate + edit生成と編集を統合
Native RGBA透明背景に対応
SIGNAL explanatory diagram · Not a product screenshotSIGNAL解説図 · 実際の製品画面ではありません
Alibaba releases open-weight Qwen-Image-2.1 image modelアリババ、画像生成・編集モデルQwen-Image-2.1を公開
4 linked sites関連4サイト
Z.ai open-sources ZCode after covert code-upload backlashZ.aiのZCode、無断アップロード発覚で公開
7 linked sites関連7サイト