Palm 2 Models by Google DeepMind
About
PaLM 2 is Google's advanced family of large language models, building on its predecessor, PaLM. It offers substantial enhancements in multilingual abilities, reasoning, and coding, tailored for key developer use cases. Trained on an extensive dataset from diverse sources such as web documents, books, and code, it excels in understanding and generating complex text across over 100 languages. With variations like Gecko for mobile and Bison and Unicorn for more demanding tasks, it provides flexibility for various applications. Integrated into products like Bard and available through the PaLM API, PaLM 2 aims to outperform earlier models and even compete with GPT-4, though independent assessments have shown varied results 123.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 8k context and structured outputs.
Use when the workload needs 8k context and structured outputs.
Use when the workload needs 8k context and structured outputs.
Use when the workload needs 8k context and structured outputs.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Text Bison | Use when the workload needs 8k context and structured outputs. | 2023-12 | 8k contextstructured outputs | Current |
| Chat Bison | Use when the workload needs 8k context and structured outputs. | 2023-12 | 8k contextstructured outputs | Current |
| PaLM 2 (text-bison) on Google Vertex AI | Use when the workload needs 8k context and structured outputs. | 2023-12 | 8k contextstructured outputs | Current |
| PaLM 2 (chat-bison) on Google Vertex AI | Use when the workload needs 8k context and structured outputs. | 2023-12 | 8k contextstructured outputs | Current |
Release Timeline
1 release groupSpecifications(4 models)
| Model | Released | Context | Structured Outputs |
|---|---|---|---|
| Text Bison | 2023-12 | 8k | Yes |
| Chat Bison | 2023-12 | 8k | Yes |
| PaLM 2 (text-bison) on Google Vertex AI | 2023-12 | 8k | Yes |
| PaLM 2 (chat-bison) on Google Vertex AI | 2023-12 | 8k | Yes |
Available From(1 provider)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| PaLM 2 (text-bison) on Google Vertex AI | GCP Vertex AI | $0.125 | $0.375 | Serverless |
| PaLM 2 (chat-bison) on Google Vertex AI | GCP Vertex AI | $0.125 | $0.375 | Serverless |
| Chat Bison | GCP Vertex AI | $0.5 | $0.5 | Serverless |
| Text Bison | GCP Vertex AI | $0.5 | $0.5 | Serverless |
Frequently Asked Questions
- What is Palm 2 used for?
- Palm 2 is used for upcoming, structured outputs, and coding. The family description and listed model capabilities point to those workloads as the best fit.
- How does Palm 2 compare to Gemma 4?
- Palm 2 by Google DeepMind is strongest where you need upcoming, while Gemma 4 by Google DeepMind is the closest related family to check for multimodal. Palm 2 has 4 listed variants and reaches up to 8k context, while Gemma 4 reaches up to 256k context, so compare the specs and pricing tables before choosing a production model.
- Which Palm 2 model should I use?
- For the lowest listed input price, start with PaLM 2 (text-bison) on Google Vertex AI through GCP Vertex AI at $0.125/1M input tokens. For the most capable/latest local choice, evaluate PaLM 2 (text-bison) on Google Vertex AI with 8k context and structured outputs.
