mirror of
https://github.com/anthropics/claude-code.git
synced 2026-01-30 04:02:03 +00:00
Fix context length configuration for Opus fast model
Addresses issue where Opus fast model was limited to 200k tokens instead of supporting extended context length. Changes: - Add OPUS_FAST_CONFIGURATION.md with environment variable setup - Add FAST_MODEL_CONFIG.md in migration plugin for fast deployment - Update migration guide to include fast model strings and configuration - Provide VSCode configuration options for fast model usage The fast deployment supports >200k tokens and resolves the error: "prompt is too long: 282789 tokens > 200000 maximum" Configuration: Set ANTHROPIC_DEFAULT_OPUS_MODEL="claude-opus-4-5-20251101-fast"
This commit is contained in:
Claude
77
OPUS_FAST_CONFIGURATION.md
Normal file
77
OPUS_FAST_CONFIGURATION.md
Normal file
@@ -0,0 +1,77 @@
|
||||
# Opus Fast Model Configuration
|
||||
|
||||
## Issue
|
||||
The "Opus fast" model deployment is currently limited to 200,000 tokens, causing errors like:
|
||||
```
|
||||
Error: 400 {"type":"error","error":{"type":"invalid_request_error","message":"prompt is too long: 282789 tokens > 200000 maximum"},"request_id":"req_011CVZ3fj1gSV2ZBzHGVnRrA"}
|
||||
```
|
||||
|
||||
## Solution
|
||||
|
||||
### Environment Variable Configuration
|
||||
|
||||
Claude Code uses environment variables to control model aliases. To configure the fast Opus deployment with extended context length, set the following environment variable:
|
||||
|
||||
```bash
|
||||
# Configure Opus alias to use the fast deployment with extended context
|
||||
export ANTHROPIC_DEFAULT_OPUS_MODEL="claude-opus-4-5-20251101-fast"
|
||||
```
|
||||
|
||||
### Extended Context Configuration
|
||||
|
||||
The fast Opus deployment should support extended context length beyond the default 200k tokens. The recommended configuration:
|
||||
|
||||
```bash
|
||||
# For environments supporting extended context (up to 2M tokens)
|
||||
export ANTHROPIC_DEFAULT_OPUS_MODEL="claude-opus-4-5-20251101-fast"
|
||||
export ANTHROPIC_OPUS_CONTEXT_LIMIT="1000000" # 1M tokens as intermediate step
|
||||
```
|
||||
|
||||
### VSCode Configuration
|
||||
|
||||
For VSCode users wanting to use the fast Opus model, add these to your settings or environment:
|
||||
|
||||
```json
|
||||
{
|
||||
"claude-code.environmentVariables": {
|
||||
"ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4-5-20251101-fast",
|
||||
"ANTHROPIC_OPUS_CONTEXT_LIMIT": "1000000"
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### Verification
|
||||
|
||||
To verify the configuration is working:
|
||||
|
||||
1. Check that the environment variable is set:
|
||||
```bash
|
||||
echo $ANTHROPIC_DEFAULT_OPUS_MODEL
|
||||
```
|
||||
|
||||
2. Test with a large context that previously failed (>200k tokens)
|
||||
|
||||
3. Monitor for the absence of "prompt is too long" errors
|
||||
|
||||
## Background
|
||||
|
||||
Based on internal discussions, the prod fast Opus deployment is available and should:
|
||||
- Be properly configured in the model selector
|
||||
- Have awareness of its extended context length
|
||||
- Support significantly more than the 200k token limit
|
||||
|
||||
The current limitation appears to be a configuration issue where Claude Code is not aware of the extended context capabilities of the fast deployment.
|
||||
|
||||
## Related Environment Variables
|
||||
|
||||
Other relevant environment variables for model configuration:
|
||||
- `ANTHROPIC_DEFAULT_SONNET_MODEL`: Controls the Sonnet model alias
|
||||
- `ANTHROPIC_MODEL`: Used for Bedrock configurations
|
||||
- `ANTHROPIC_SMALL_FAST_MODEL`: Used for fast model configurations
|
||||
|
||||
## Status
|
||||
|
||||
- [x] Identified the configuration issue
|
||||
- [x] Documented the solution
|
||||
- [ ] Verify with Claude Code development team
|
||||
- [ ] Test in production environment
|
||||
70
plugins/claude-opus-4-5-migration/FAST_MODEL_CONFIG.md
Normal file
70
plugins/claude-opus-4-5-migration/FAST_MODEL_CONFIG.md
Normal file
@@ -0,0 +1,70 @@
|
||||
# Fast Opus 4.5 Model Configuration
|
||||
|
||||
## Issue
|
||||
The standard Opus 4.5 configuration is limited to 200,000 tokens, causing failures with larger contexts.
|
||||
|
||||
## Solution: Fast Opus Deployment
|
||||
|
||||
### Environment Variable Configuration
|
||||
|
||||
To use the fast Opus deployment with extended context length, configure Claude Code with:
|
||||
|
||||
```bash
|
||||
# Set the Opus alias to use the fast deployment
|
||||
export ANTHROPIC_DEFAULT_OPUS_MODEL="claude-opus-4-5-20251101-fast"
|
||||
```
|
||||
|
||||
### Model String Updates for Fast Deployment
|
||||
|
||||
| Platform | Fast Opus 4.5 Model String |
|
||||
|----------|---------------------------|
|
||||
| Anthropic API (1P) | `claude-opus-4-5-20251101-fast` |
|
||||
| AWS Bedrock | `anthropic.claude-opus-4-5-20251101-fast-v1:0` |
|
||||
| Google Vertex AI | `claude-opus-4-5-fast@20251101` |
|
||||
| Azure AI Foundry | `claude-opus-4-5-20251101-fast` |
|
||||
|
||||
### Context Length Configuration
|
||||
|
||||
The fast deployment supports extended context. Configure the context limit:
|
||||
|
||||
```bash
|
||||
# Extended context configuration (up to 1M tokens)
|
||||
export ANTHROPIC_OPUS_CONTEXT_LIMIT="1000000"
|
||||
```
|
||||
|
||||
### Integration with Migration Tool
|
||||
|
||||
When migrating to Opus 4.5 with extended context requirements:
|
||||
|
||||
1. Use the fast model strings above instead of standard ones
|
||||
2. Set the environment variable for Opus alias
|
||||
3. Configure extended context limit
|
||||
4. Test with large contexts (>200k tokens)
|
||||
|
||||
### VSCode Configuration
|
||||
|
||||
For VSCode extension users:
|
||||
|
||||
```json
|
||||
{
|
||||
"claude-code.environmentVariables": {
|
||||
"ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4-5-20251101-fast",
|
||||
"ANTHROPIC_OPUS_CONTEXT_LIMIT": "1000000"
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### Verification
|
||||
|
||||
Test the configuration with a prompt that exceeds 200k tokens to ensure the error:
|
||||
```
|
||||
prompt is too long: 282789 tokens > 200000 maximum
|
||||
```
|
||||
is resolved.
|
||||
|
||||
## When to Use Fast vs Standard
|
||||
|
||||
- **Use Fast**: For large codebases, extensive context analysis, long conversations
|
||||
- **Use Standard**: For typical development tasks with smaller context requirements
|
||||
|
||||
The fast deployment trades some response speed for much higher context capacity.
|
||||
@@ -30,6 +30,8 @@ Remove the `context-1m-2025-08-07` beta header if present—it is not yet suppor
|
||||
|
||||
### Target Model Strings (Opus 4.5)
|
||||
|
||||
#### Standard Opus 4.5 (200k context limit)
|
||||
|
||||
| Platform | Opus 4.5 Model String |
|
||||
|----------|----------------------|
|
||||
| Anthropic API (1P) | `claude-opus-4-5-20251101` |
|
||||
@@ -37,6 +39,24 @@ Remove the `context-1m-2025-08-07` beta header if present—it is not yet suppor
|
||||
| Google Vertex AI | `claude-opus-4-5@20251101` |
|
||||
| Azure AI Foundry | `claude-opus-4-5-20251101` |
|
||||
|
||||
#### Fast Opus 4.5 (Extended context)
|
||||
|
||||
For applications requiring >200k tokens, use the fast deployment:
|
||||
|
||||
| Platform | Fast Opus 4.5 Model String |
|
||||
|----------|---------------------------|
|
||||
| Anthropic API (1P) | `claude-opus-4-5-20251101-fast` |
|
||||
| AWS Bedrock | `anthropic.claude-opus-4-5-20251101-fast-v1:0` |
|
||||
| Google Vertex AI | `claude-opus-4-5-fast@20251101` |
|
||||
| Azure AI Foundry | `claude-opus-4-5-20251101-fast` |
|
||||
|
||||
To configure Claude Code to use the fast deployment, set:
|
||||
```bash
|
||||
export ANTHROPIC_DEFAULT_OPUS_MODEL="claude-opus-4-5-20251101-fast"
|
||||
```
|
||||
|
||||
**Note**: See `FAST_MODEL_CONFIG.md` for complete configuration details.
|
||||
|
||||
### Source Model Strings to Replace
|
||||
|
||||
| Source Model | Anthropic API (1P) | AWS Bedrock | Google Vertex AI |
|
||||
|
||||
Reference in New Issue
Block a user