Skip to content

Commit 45c2d91

Browse files
jsamuel1awsarron
authored andcommitted
feat: upgrade to Claude Sonnet 4 model (strands-agents#14)
- Updated default model from claude-3-7-sonnet to claude-sonnet-4-20250514-v1:0 - Updated model_id in .prompt, README.md, and model_utils.py - Adjusted max_tokens to 32768 for Sonnet 4 compatibility - Updated documentation references to reflect latest model capabilities This upgrade provides enhanced reasoning capabilities and performance improvements.
1 parent 3814635 commit 45c2d91

File tree

3 files changed

+9
-9
lines changed

3 files changed

+9
-9
lines changed

.prompt

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -182,7 +182,7 @@ from botocore.config import Config
182182

183183
# Configure Bedrock model with maximum capabilities
184184
model = BedrockModel(
185-
model_id="us.anthropic.claude-3-7-sonnet-20250219-v1:0",
185+
model_id="us.anthropic.claude-sonnet-4-20250514-v1:0",
186186
max_tokens=int(os.getenv("STRANDS_MAX_TOKENS", "64000")), # Maximized token output
187187
boto_client_config=Config(
188188
read_timeout=900, # 15 min timeout for long operations
@@ -378,9 +378,9 @@ export STRANDS_SYSTEM_PROMPT="Your custom prompt here"
378378
- Create agents that can collaborate through shared interfaces
379379

380380
6. **Maximizing Bedrock Performance**
381-
- Use Claude 3.7 Sonnet for best reasoning capabilities
381+
- Use the latest Claude Sonnet model available for best reasoning capabilities
382382
- Configure extended timeouts for complex reasoning tasks
383383
- Enable thinking capability with appropriate token budget
384384
- Use maximal output tokens to handle complex tool responses
385385

386-
I'm here to help you build self-extending agents that continuously evolve their capabilities through autonomous tool creation. Let's build something extraordinary together!
386+
I'm here to help you build self-extending agents that continuously evolve their capabilities through autonomous tool creation. Let's build something extraordinary together!

README.md

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -130,8 +130,8 @@ Strands comes with optimized, maxed-out configuration settings for the Bedrock m
130130

131131
```json
132132
{
133-
"model_id": "us.anthropic.claude-3-7-sonnet-20250219-v1:0",
134-
"max_tokens": 64000,
133+
"model_id": "us.anthropic.claude-sonnet-4-20250514-v1:0",
134+
"max_tokens": 32767,
135135
"boto_client_config": {
136136
"read_timeout": 900,
137137
"connect_timeout": 900,
@@ -150,8 +150,8 @@ Strands comes with optimized, maxed-out configuration settings for the Bedrock m
150150
```
151151

152152
These settings provide:
153-
- Claude 3.7 Sonnet (latest high-performance model)
154-
- Maximum token output (64,000 tokens)
153+
- Claude Sonnet 4 (latest high-performance model)
154+
- Maximum token output (32,768 tokens)
155155
- Extended timeouts (15 minutes) for complex operations
156156
- Automatic retries with adaptive backoff
157157
- Enabled thinking capability with 2,048 token budget for recursive reasoning

src/strands_agents_builder/utils/model_utils.py

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -11,8 +11,8 @@
1111

1212
# Default model configuration
1313
DEFAULT_MODEL_CONFIG = {
14-
"model_id": "us.anthropic.claude-3-7-sonnet-20250219-v1:0",
15-
"max_tokens": int(os.getenv("STRANDS_MAX_TOKENS", "64000")),
14+
"model_id": "us.anthropic.claude-sonnet-4-20250514-v1:0",
15+
"max_tokens": int(os.getenv("STRANDS_MAX_TOKENS", "32768")),
1616
"boto_client_config": Config(
1717
read_timeout=900,
1818
connect_timeout=900,

0 commit comments

Comments
 (0)