Pure state-space language model from TII, announced August 2024. 7 billion parameters using an attention-free SSLM architecture capable of processing arbitrary-length sequences in constant memory (demonstrated generation of 130,000+ tokens without memory scaling). Trained on approximately 5.5 trillion tokens primarily from RefinedWeb with technical and code data from public sources and a curated final-stage mix. TII reports average benchmark scores of 64.09 across ARC, HellaSwag, MMLU, Winogrande, TruthfulQA, and GSM8K, outperforming Mistral-7B-v0.1 (60.97) and matching gemma-7B (63.75). Distributed under the TII Falcon Mamba 7B License 1.0.
信号概览
评分详解
| 信号 | 强度 | 权重 | 影响 |
|---|---|---|---|
| Pricingjust now | 100 | 25% | +25.0 |
| Output Capacityjust now | 65 | 15% | +9.8 |
| Context Windowjust now | 65 | 15% | +9.7 |
| Capabilitiesjust now | 17 | 30% | +5.0 |
| Recencyjust now | 6 | 15% | +0.9 |
功能
模态
TII近期发布时间线
把当前模型放回同一服务商最近的发布节奏中查看。
Falcon-H1-Arabic 34B Instruct
coding
Falcon-H1-Arabic 7B Instruct
coding
Falcon-H1-Arabic 3B Instruct
coding
Falcon Arabic 7B Instruct
coding
Falcon3 10B Instruct
coding
Falcon3 7B Instruct
coding
Falcon Mamba 7B Instruct当前模型
coding
评测
社区和从业者反馈在基准测试和价格之上增加了真实世界的信号。
Reviews
Be the first to review this model
Share your experience with Falcon Mamba 7B Instruct and help the community make better decisions.
常见问题
关键信息
价格工具
访问与可用性
价格
免费