Falcon Models by Technology Innovation Institute (TII)
About
The Falcon family of large language models (LLMs), developed by the Technology Innovation Institute (TII) in Abu Dhabi, offers a diverse range of models that are open-source and freely available for research and commercial applications. Notably, this includes the Falcon-40B, which excelled on the Hugging Face Open LLM Leaderboard, and the even more advanced Falcon-180B, which matches the performance of many proprietary models. These models benefit from training on extensive datasets like the RefinedWeb dataset, known for its high-quality, filtered, and deduplicated web content. Additionally, the Falcon family includes instruction-tuned versions, such as Falcon-7B-Instruct and Falcon-40B-Instruct, optimized for conversational interactions. Recently, the Falcon Mamba 7B was introduced, offering improved memory efficiency and enhanced long-text generation through its novel state-space language model (SSLM) architecture. This family of models underscores a strong commitment to open-source AI, making advanced language capabilities accessible to a broad audience of researchers and users 1481011.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 128k context and 7B parameters.
Use when the workload needs 40B parameters and structured outputs.
Use when the workload needs 7B parameters and structured outputs.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Falcon 3 7B Instruct | Use when the workload needs 128k context and 7B parameters. | 2024-12 | 128k context7B parameters | Current |
| Falcon 180B | Use when the workload needs 180B parameters. | 2023-11 | 180B parameters | Current |
| Falcon 40B | Use when the workload needs 40B parameters and structured outputs. | 2023-11 | 40B parametersstructured outputs | Current |
| Falcon 7B | Use when the workload needs 7B parameters and structured outputs. | 2023-11 | 7B parametersstructured outputs | Current |
Release Timeline
2 release groupsSpecifications(4 models)
| Model | Released | Context | Parameters | Structured Outputs |
|---|---|---|---|---|
| Falcon 3 7B Instruct | 2024-12 | 128k | 7B | No |
| Falcon 180B | 2023-11 | — | 180B | No |
| Falcon 40B | 2023-11 | — | 40B | Yes |
| Falcon 7B | 2023-11 | — | 7B | Yes |
Available From(6 providers)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| Falcon 7B | Microsoft Foundry | $0.52 | $0.67 | Provisioned |
| Falcon 40B | Replicate API | $0.65 | $2.75 | Serverless |
| Falcon 40B | Microsoft Foundry | $1.54 | $1.77 | Provisioned |
Frequently Asked Questions
- What is Falcon used for?
- Falcon is used for structured outputs, coding, and long-context generation. The family description and listed model capabilities point to those workloads as the best fit.
- How does Falcon compare to MOSS-Audio?
- Falcon by Technology Innovation Institute (TII) is strongest where you need structured outputs, while MOSS-Audio by MOSI Intelligence is the closest related family to check for multimodal. Falcon has 4 listed variants and reaches up to 128k context, so compare the specs and pricing tables before choosing a production model.
- Which Falcon model should I use?
- For the lowest listed input price, start with Falcon 7B through Microsoft Foundry at $0.52/1M input tokens. For the most capable/latest local choice, evaluate Falcon 40B with structured outputs.
