Skip to content

Commit 058acc9

Browse files
committed
feat: add OpenAI GPT-5.5 and GPT-5.4 via Bedrock Mantle
- Add gpt-5.5 and gpt-5.4 models to configmap model_list - Models use Responses API via bedrock-mantle.us-east-2.api.aws endpoint - Auth via BEDROCK_MANTLE_API_KEY (bearer token) or SigV4 - Update docs/models.md with Bedrock Mantle section - Tested: both models respond correctly via litellm.responses() Ref: https://aws.amazon.com/blogs/aws/get-started-with-openai-gpt-5-5-gpt-5-4-models-and-codex-on-amazon-bedrock/
1 parent 5246a75 commit 058acc9

2 files changed

Lines changed: 51 additions & 0 deletions

File tree

docs/models.md

Lines changed: 33 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -20,11 +20,44 @@
2020
| `claude-sonnet-3-7` | `us.anthropic.claude-3-7-sonnet-20250219-v1:0` | us | |
2121
| `claude-haiku-4-5` | `global.anthropic.claude-haiku-4-5-20251001-v1:0` | global | |
2222
| `bedrock/*` | 任意 Bedrock 模型 ID | 通配符(直接透传)| |
23+
| `gpt-5.5` | `openai.gpt-5.5` | us-east-2 (Mantle) | **** OpenAI GPT-5.5 via Bedrock Mantle |
24+
| `gpt-5.4` | `openai.gpt-5.4` | us-east-2 (Mantle) | **** OpenAI GPT-5.4 via Bedrock Mantle |
2325

2426
**区域类型说明**:
2527
- **us** - Cross-Region Inference 端点,us-west-2 和 us-east-1 自动负载均衡
2628
- **global** - Global Inference 端点,跨多个区域全球负载均衡
2729

30+
31+
32+
---
33+
34+
## OpenAI GPT 模型(Bedrock Mantle)
35+
36+
2026-06-01 GA,OpenAI GPT-5.5 和 GPT-5.4 通过 Amazon Bedrock Mantle 提供服务。
37+
38+
| 模型名 | Model ID | 特点 |
39+
|--------|----------|------|
40+
| `gpt-5.5` | `openai.gpt-5.5` | 最强推理能力,支持 reasoning (high/medium/low) |
41+
| `gpt-5.4` | `openai.gpt-5.4` | 性价比优选,适合日常任务 |
42+
43+
**注意事项**
44+
- 这些模型仅支持 **Responses API** (`/v1/responses`),不支持 `/chat/completions`
45+
- Endpoint: `https://bedrock-mantle.us-east-2.api.aws/openai/v1/responses`
46+
- 认证方式:`BEDROCK_MANTLE_API_KEY`(Bearer Token)或 AWS SigV4
47+
- 不支持的 tool types:`computer_use_preview``web_search`
48+
- 支持的 tool types:`function``mcp``custom``namespace``tool_search`
49+
50+
**通过 LiteLLM 调用**
51+
```python
52+
import litellm
53+
# 使用 Responses API
54+
response = litellm.responses(
55+
model="openai/openai.gpt-5.5",
56+
input="Your prompt here",
57+
)
58+
```
59+
60+
**参考**[AWS Blog - Get started with OpenAI GPT-5.5, GPT-5.4 models, and Codex on Amazon Bedrock](https://aws.amazon.com/blogs/aws/get-started-with-openai-gpt-5-5-gpt-5-4-models-and-codex-on-amazon-bedrock/)
2861
---
2962

3063
## Fallback 降级链

kubernetes/configmap.yaml

Lines changed: 18 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -216,6 +216,24 @@ data:
216216
model_info:
217217
id: alias-us-anthropic-opus-4-1
218218
model_name: us.anthropic.claude-opus-4-1-20250805-v1:0
219+
# OpenAI GPT-5.5 / GPT-5.4 on Bedrock Mantle (2026-06-01 GA)
220+
# These models use the Responses API (/v1/responses) only.
221+
# Auth: SigV4 via IRSA or BEDROCK_MANTLE_API_KEY bearer token.
222+
# Region: us-east-2 (launch region for GPT-5.5/5.4)
223+
- litellm_params:
224+
model: openai/openai.gpt-5.5
225+
api_base: https://bedrock-mantle.us-east-2.api.aws/openai/v1
226+
api_key: os.environ/BEDROCK_MANTLE_API_KEY
227+
model_info:
228+
id: gpt-5-5
229+
model_name: gpt-5.5
230+
- litellm_params:
231+
model: openai/openai.gpt-5.4
232+
api_base: https://bedrock-mantle.us-east-2.api.aws/openai/v1
233+
api_key: os.environ/BEDROCK_MANTLE_API_KEY
234+
model_info:
235+
id: gpt-5-4
236+
model_name: gpt-5.4
219237
router_settings:
220238
allowed_fails: 3
221239
cooldown_time: 60

0 commit comments

Comments
 (0)