Largest model in the general-purpose Falcon 3 family released by TII in December 2024. 10 billion parameters with a decoder-only Llama-compatible transformer architecture (LlamaForCausalLM), SwiGLU activation, GQA with 12 attention heads and 4 key-value heads, 131,072 vocabulary size, and 32,768 position embeddings. Trained on 14 trillion tokens followed by 2 trillion tokens of continued training on diverse data. TII reports MMLU 73.1, GSM8K 83.0, MBPP 73.8, BBH 59.7, BFCL 86.3, and MMLU-PRO 42.5 on the 10B variant. Distributed on Hugging Face under the Falcon LLM License (Apache-2.0-based with TII-specific terms).
| 信号 | 强度 | 权重 | 影响 |
|---|---|---|---|
| Pricingjust now | 100 | 25% | +25.0 |
| Capabilitiesjust now | 50 | 30% | +15.0 |
| Context Windowjust now | 72 | 15% | +10.7 |
| Output Capacityjust now | 65 | 15% | +9.8 |
| Recencyjust now | 46 | 15% | +6.8 |
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当前模型
coding
Falcon3 7B Instruct
coding
Falcon Mamba 7B Instruct
coding
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。
Share your experience with Falcon3 10B Instruct and help the community make better decisions.