This page gives you information about legacy generative AI models on Vertex AI. The models in a legacy model family are no longer updated with new stable versions. After all stable versions of a legacy model reach their discontinuation dates, the model family is no longer supported.
PaLM API models
The following table summarizes PaLM API legacy models:
| Model name | Description | Model properties | Tuning support |
|---|---|---|---|
| PaLM 2 for Text ( text-bison) |
Fine-tuned to follow natural language instructions and is suitable for a variety of language tasks, such as classification, summarization, and extraction. | Maximum input tokens: 8192 Maximum output tokens: 1024 Training data: Up to Feb 2023 |
Supervised: Yes RLHF: Yes (Preview) Distillation: No |
PaLM 2 for Text (text-unicorn) |
The most advanced text model in the PaLM family of models for use with complex natural language tasks. | Maximum input tokens: 8192 Maximum output tokens: 1024 Training data: Up to Feb 2023 |
Supervised: No RLHF: No Distillation: Yes (Preview) |
| PaLM 2 for Text 32k ( text-bison-32k) |
Fine-tuned to follow natural language instructions and is suitable for a variety of language tasks. | Max tokens (input + output): 32,768 Max output tokens: 8,192 Training data: Up to Aug 2023 |
Supervised: Yes RLHF: No Distillation: No |
| PaLM 2 for Chat ( chat-bison) |
Fine-tuned for multi-turn conversation use cases. | Maximum input tokens: 8192 Maximum output tokens: 2048 Training data: Up to Feb 2023 Maximum turns : 2500 |
Supervised: Yes RLHF: No Distillation: No |
| PaLM 2 for Chat 32k ( chat-bison-32k) |
Fine-tuned for multi-turn conversation use cases. | Max tokens (input + output): 32,768 Max output tokens: 8,192 Training data: Up to Aug 2023 Max turns : 2500 |
Supervised: Yes RLHF: No Distillation: No |
Codey APIs models
The following table summarizes Codey APIs legacy models. Note that
code-gecko code completion model is not a legacy model.
| Model name | Description | Model properties | Tuning support |
|---|---|---|---|
| Codey for Code Generation ( code-bison) |
A model fine-tuned to generate code based on a natural language description of the desired code. For example, it can generate a unit test for a function. | Maximum input tokens: 6144 Maximum output tokens: 1024 |
Supervised: Yes RLHF: No Distillation: No |
| Codey for Code Generation 32k ( code-bison-32k) |
A model fine-tuned to generate code based on a natural language description of the desired code. For example, it can generate a unit test for a function. | Max tokens (input + output): 32,768 Max output tokens: 8,192 |
Supervised: Yes RLHF: No Distillation: No |
| Codey for Code Chat ( codechat-bison) |
A model fine-tuned for chatbot conversations that help with code-related questions. | Maximum input tokens: 6144 Maximum output tokens: 1024 |
Supervised: Yes RLHF: No Distillation: No |
| Codey for Code Chat 32k ( codechat-bison-32k) |
A model fine-tuned for chatbot conversations that help with code-related questions. | Max tokens (input + output): 32,768 Max output tokens: 8,192 |
Supervised: Yes RLHF: No Distillation: No |
Language support
Vertex AI PaLM API and Codey APIs legacy models support the following languages:
- Arabic (
ar) - Bengali (
bn) - Bulgarian (
bg) - Chinese simplified and traditional (
zh) - Croatian (
hr) - Czech (
cs) - Danish (
da) - Dutch (
nl) - English (
en) - Estonian (
et) - Finnish (
fi) - French (
fr) - German (
de) - Greek (
el) - Hebrew (
iw) - Hindi (
hi) - Hungarian (
hu) - Indonesian (
id) - Italian (
it) - Japanese (
ja) - Korean (
ko) - Latvian (
lv) - Lithuanian (
lt) - Norwegian (
no) - Polish (
pl) - Portuguese (
pt) - Romanian (
ro) - Russian (
ru) - Serbian (
sr) - Slovak (
sk) - Slovenian (
sl) - Spanish (
es) - Swahili (
sw) - Swedish (
sv) - Thai (
th) - Turkish (
tr) - Ukrainian (
uk) - Vietnamese (
vi)
Legacy model discontinuation date
The following table shows the discontinuation date of legacy models:
| chat-bison model | Release date | Discontinuation date |
|---|---|---|
| chat-bison@002 | December 6, 2023 | April 9, 2025 |
| chat-bison-32k model | Release date | Discontinuation date |
|---|---|---|
| chat-bison-32k@002 | December 4, 2023 | April 9, 2025 |
| code-bison model | Release date | Discontinuation date |
|---|---|---|
| code-bison@002 | December 6, 2023 | April 9, 2025 |
| code-bison-32k model | Release date | Discontinuation date |
|---|---|---|
| code-bison-32k@002 | December 4, 2023 | April 9, 2025 |
| codechat-bison model | Release date | Discontinuation date |
|---|---|---|
| codechat-bison@002 | December 6, 2023 | April 9, 2025 |
| codechat-bison-32k model | Release date | Discontinuation date |
|---|---|---|
| codechat-bison-32k@002 | December 4, 2023 | April 9, 2025 |
| text-bison model | Release date | Discontinuation date |
|---|---|---|
| text-bison@002 | December 6, 2023 | April 9, 2025 |
| text-bison-32k model | Release date | Discontinuation date |
|---|---|---|
| text-bison-32k@002 | December 4, 2023 | April 9, 2025 |
| text-unicorn model | Release date | Discontinuation date |
|---|---|---|
| text-unicorn@001 | November 30, 2023 | April 9, 2025 |
Legacy models that support Provisioned Throughput
This table shows legacy models that support Provisioned Throughput, which is measured in characters per second, minimum purchase increments, and burndown rates.
| Model | Throughput per GSU | Minimum GSU purchase increment | Burndown rates |
|---|---|---|---|
text-bison, chat-bison,
code-bison, codechat-bison |
4,000 | 1 | 1 input char = 1 char 1 output char = 2 chars |
text-unicorn |
400 | 1 | 1 input char = 1 char 1 output char = 3 chars |