| Spesifikasi | Nilai |
|---|---|
| Pembuat | IBM |
| Model | H 1B (1.5B), H 350M (340M), transformer 1B dan 350M |
| Panjang jujukan | 128K (1B) / 32K (350M) token |
| Lesen | Apache 2.0 |
Daripada kad IBM di Hugging Face, yang jadual penilaian oleh IBM menjadi sumber carta; pemberat terbuka, jadi tiada harga API. · sumber
Estimate from the version's published specs, benchmarks and reviews, 0-5; scores are only comparable within one criteria year. Scored on the flagship Granite 4.0 H 1B from IBM's Hugging Face model card (opened this run). Context 128K tokens (the 350M models are 32K) scores 1; text-only input scores 1. Reasoning, coding, price and speed are null: no Artificial Analysis Intelligence Index entry or Terminal-Bench figure is published (IBM gives only MMLU-Pro 32.9 and GPQA 29.7, which the anchors cannot map), the open weights carry no list price, and no output speed is published. Three independent press pages were found (SiliconANGLE, It's FOSS, Open Source For You); all are launch coverage restating IBM's own benchmarks, not hands-on tests, and searches for hands-on reviews of the 1B/350M models found none.[2]
The tradeoff is a reduction in accuracy and contextual knowledge that is trimmed from the models to reduce their size.
The results show Granite Nano outperforming several similarly sized models when averaged across these domains.
Performance benchmarks show Granite 4.0 H 1B leading in instruction-following and function-calling tasks, achieving 78.5 on IFEval and 54.8 on Berkeley’s Function Calling Leaderboard v3, outperforming Qwen3 1.7B and Gemma 3 1B.
| IBM releases small open-source Granite 4 models for mobile devices and browsers - SiliconANGLE · 29 Okt 2025 | 0.10 |
|---|---|
| IBM Releases Its Smallest AI Models Yet (Available via Hugging Face) · 29 Okt 2025 | 0.50 |
| IBM Launches Granite 4 Nano Under Apache 2.0 For Local Generative AI - Open Source For You · 4 Nov 2025 | 0.60 |
Foto daripada pembaca, berasingan daripada media bersumber pin. Nilaikan apa yang anda makan atau lihat, dan pada pin restoran, tag hidangan terus pada foto anda.
Entri pertama sentiasa sumber pin. Keyakinan keseluruhan ialah purata berpemberat bagi sejauh mana setiap rujukan menyokong masa mula dan tamat yang digunakan di atas; sesebuah rujukan diambil kira separuh kurang bagi setiap 180 hari lebih lama daripada yang terbaharu.
Catatan blog IBM di Hugging Face[2][3] 'Granite 4.0 Nano: Just how small can you go?' serta kad model granite-4.0-h-1b dan granite-4.0-h-350m menyatakan 28 Oktober 2025; It's FOSS melaporkannya pada hari berikutnya.
Model card of the hybrid 1B Nano model: Apache 2.0 and 'Release Date: October 28, 2025'.[1]
Model card of the hybrid 350M Nano model, same release date.[1]
Ada yang tertinggal atau salah? Nyatakan dengan kata-kata anda sendiri: pautan yang menyokong pin ini, tarikh mula atau tamat yang berbeza beserta sebabnya, atau fakta yang tiada atau tersilap. AI menyemaknya berbanding sumber pin ini, mencari yang lebih baik, dan menambah mana-mana halaman yang menyokong anda. Sumber pin itu sendiri tetap paling diambil kira. Gambar yang menunjukkan sesuatu yang lain, atau menunjukkannya dengan buruk, turut dilihat, dan dialihkan ke bawah atau diganti.
Tajuk asal: Granite 4.0 Nano 1B and 350M Released