3 methods to solve model not found issues when configuring Claude Code with third-party API proxy services

title: Claude Code "Model Not Found" Error? Here's the Complete Fix for API Proxies
description: Fix Claude Code "model not found" errors when using third-party API proxy services like APIYI. Learn correct Base URL and settings.json configurations.
tags: [Claude Code, API Proxy, APIYI, Troubleshooting, Large Language Model]

Author's Note: A complete solution for the "model not found" error when connecting Claude Code to third-party API proxy services. Includes correct Base URL syntax, settings.json configuration methods, and the latest model name reference table.

When using Claude Code, the most common error you'll likely run into is this one:

There's an issue with the selected model (claude-sonnet-4-6). It may not exist or you may not have access to it. Run /model to pick a different model.

This error typically occurs in two scenarios:

Using an official API key but with a typo in the model name (just run /model to pick a new one).
Connecting via a third-party API proxy service, where the Base URL isn't configured correctly, causing model routing to fail.

This article focuses on the complete solution for the second scenario. We'll use APIYI (apiyi.com) as our example, covering environment variables, settings.json configurations, and the latest Claude model name reference table.

Core Value: By the end of this post, you'll know exactly how to configure Claude Code to work with a third-party API proxy service, eliminating "model not found" errors and ensuring stable access to all Claude Code features.

I. Quick Diagnosis: Is it the Model Name or the Base URL?

Before you start tweaking settings, take 30 seconds to figure out which issue you're facing:

Symptom	Likely Cause	Solution Path
Error: `Run /model to pick a different model`	Model name doesn't exist or no permission	Run `/model` to reselect
Correct model name entered, but still errors	Base URL configuration issue	Check `ANTHROPIC_BASE_URL`
API key validation failed	Key is invalid or not set	Reconfigure `ANTHROPIC_API_KEY`
Network timeout / Connection refused	Incorrect proxy address	Check Base URL format
All models failing	Proxy doesn't support Anthropic format	Confirm proxy supports native Anthropic protocol

If you're using an official Anthropic API key and can access anthropic.com normally, simply run the /model command within Claude Code to switch models—no extra configuration needed.

If you're using a third-party API proxy service (like apiyi.com), please continue with the configuration steps below.

🎯 Pro Tip: We recommend using the APIYI (apiyi.com) platform to access Claude Code. It's fully compatible with the native Anthropic format and supports the entire model lineup, including claude-3-5-sonnet-20241022 and claude-3-opus-20240229, with stable connectivity.

2. Core Configuration for Connecting Third-Party Proxy APIs to Claude Code

2.1 Correct Base URL Format: Drop the /v1

This is the most critical step and the place where people most often trip up.

Claude Code has specific logic for handling the Base URL: It automatically appends /v1/messages to whatever Base URL you've set. Therefore:

❌ Incorrect Format: ANTHROPIC_BASE_URL=https://api.apiyi.com/v1
- The actual request path becomes: https://api.apiyi.com/v1/v1/messages (doubling up on the /v1)
- Result: The route can't find the endpoint, and the model throws an error.
✅ Correct Format: ANTHROPIC_BASE_URL=https://api.apiyi.com
- The actual request path becomes: https://api.apiyi.com/v1/messages
- Result: Successfully hits the native Anthropic interface.

📌 Summary: When setting ANTHROPIC_BASE_URL, just use the root domain path. Don't add the /v1 suffix. Claude Code will handle the rest of the path automatically.

2.2 Getting Your API Key

Token Selection Recommendations:

Token Type	Best For	Discount
Default Token	General use, works out of the box	Standard Price
ClaudeCode Group Token	Specifically for Claude Code	12% Off (0.88x)

When creating a new token, selecting the ClaudeCode group gives you a 12% discount, making it much more cost-effective for long-term use.

III. Detailed Guide to the Two Configuration Methods

Method 1: Environment Variable Configuration (Recommended for Temporary Testing)

Run the following commands in your terminal. They'll take effect immediately for the current session:

# Set the third-party API proxy service Base URL (Note: don't add /v1)
export ANTHROPIC_BASE_URL="https://api.apiyi.com"

# Set your API key (replace with your actual key)
export ANTHROPIC_API_KEY="sk-xxxxxxxxxxxxxxxxxxxxxxxx"

# Launch Claude Code
claude

Verify if the configuration is active:

# Check current environment variables
echo $ANTHROPIC_BASE_URL
echo $ANTHROPIC_API_KEY

# You should see:
# https://api.apiyi.com
# sk-xxxxxx...

Pros & Cons:

✅ Simple and fast, no file editing needed.
✅ Doesn't affect other session configurations.
❌ Needs to be reset every time you open a new terminal (unless added to .zshrc / .bashrc).

Permanent Solution (Write to your Shell config file):

# For zsh users (macOS default)
echo 'export ANTHROPIC_BASE_URL="https://api.apiyi.com"' >> ~/.zshrc
echo 'export ANTHROPIC_API_KEY="sk-your-key"' >> ~/.zshrc
source ~/.zshrc

# For bash users
echo 'export ANTHROPIC_BASE_URL="https://api.apiyi.com"' >> ~/.bashrc
echo 'export ANTHROPIC_API_KEY="sk-your-key"' >> ~/.bashrc
source ~/.bashrc

Method 2: settings.json Configuration (Recommended for Long-term Use)

Edit the ~/.claude/settings.json file (this is a global configuration that applies to all projects):

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.apiyi.com",
    "ANTHROPIC_API_KEY": "sk-your-API-key"
  }
}

🎯 Pro Tip: Use ~/.claude/settings.json for global configuration. Set it once and it works forever without needing to mess with environment variables every time. Visit APIYI (apiyi.com) to get your key; choose the ClaudeCode group tokens to enjoy a 12% discount.

If the file doesn't exist, create it manually:

# Create the .claude directory (if it doesn't exist)
mkdir -p ~/.claude

# Create settings.json
cat > ~/.claude/settings.json << 'EOF'
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.apiyi.com",
    "ANTHROPIC_API_KEY": "sk-replace-with-your-key"
  }
}
EOF

Configuration Priority (From highest to lowest):

Priority	Source	Description
1 (Highest)	Environment Variables	Exported directly in the terminal
2	`.claude/settings.local.json`	Local project config (not committed to git)
3	`.claude/settings.json`	Shared project config
4 (Lowest)	`~/.claude/settings.json`	Global user config

For the same setting, higher priority sources will override lower ones. If you have a settings.json in your project directory, keep this hierarchy in mind.

IV. Latest Model Names Supported by Claude Code

Opus 4.6 claude-opus-4-6 Strongest capability

Comprehensive capability 95%

Reasoning depth 93%

Response speed 47%

Cost-effectiveness 35%

Complex code · Deep analysis · Long documents

Recommended first choice Sonnet 4.6 claude-sonnet-4-6 The best balance between capability and speed

Comprehensive capability 83%

Reasoning depth 75%

Response speed 80%

Cost-effectiveness 75%

Daily programming · Code review · Document writing

Haiku 4.5 claude-haiku-4-5-20251001 Ultra-fast response

Comprehensive capability 60%

Reasoning depth 51%

Response speed 97%

Cost-effectiveness 94%

Fast Completion · Simple Q&A · Batch Tasks

🎯 Claude Code Scenario Selection Recommendations Daily use → Sonnet 4.6 (Recommended) | Complex architecture → Opus 4.6 | High-frequency completion → Haiku 4.5 All the above models can be invoked via the APIYI apiyi.com unified interface, with one API key to access the entire series.

These are the latest Claude model names for 2026. Make sure to enter them exactly as shown, including case and numbers:

Standard Reasoning Models

Model Name	Series	Capability	Use Case
`claude-opus-4-6`	Claude 4.6	Most Powerful	Complex code, deep analysis, long document processing
`claude-sonnet-4-6`	Claude 4.6	Balanced	Daily programming, code reviews, technical writing
`claude-haiku-4-5-20251001`	Claude 4.5	Ultra-fast	Simple Q&A, code completion, quick tasks

Enhanced Reasoning Models (Thinking Mode)

By adding the -thinking suffix to the model name, you can enable extended thinking mode, where the model performs deep reasoning before responding:

Model Name	Base Model	Features
`claude-opus-4-6-thinking`	claude-opus-4-6	Strongest reasoning; ideal for math, logic, and complex decisions
`claude-sonnet-4-6-thinking`	claude-sonnet-4-6	Balanced reasoning and speed; the go-to for daily use
`claude-haiku-4-5-20251001-thinking`	claude-haiku-4-5-20251001	Lightweight reasoning

💡 Selection Advice: For daily Claude Code use, we recommend claude-sonnet-4-6—it offers the best balance between coding quality and response speed. For complex architecture design or debugging stubborn bugs, you can switch to claude-opus-4-6 or claude-sonnet-4-6-thinking.

5. Switching Models in Claude Code

Once you've configured your Base URL and API key, there are two ways to switch models:

5.1 Using the /model Command (The Easiest Way)

In a Claude Code conversation, simply type:

/model

Claude Code will pop up a list of available models. If you're using a third-party API proxy service, the list might not automatically show every model, so you may need to type the model name manually.

5.2 Specifying Default Models in settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.apiyi.com",
    "ANTHROPIC_API_KEY": "sk-your-API-key",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-haiku-4-5-20251001",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4-6"
  }
}

By using the ANTHROPIC_DEFAULT_*_MODEL environment variables, you can specify the exact model name for each tier. This prevents Claude Code from using its built-in default names, which might not match the naming convention of your API proxy service.

🎯 Full Configuration Example: We recommend saving the above configuration in ~/.claude/settings.json. Use a ClaudeCode group token from APIYI (apiyi.com) as your ANTHROPIC_API_KEY.

6. Common Errors and Solutions (FAQ)

Q1: Claude Code won't start at all after I configured ANTHROPIC_BASE_URL?

Check if your JSON format is correct. Common mistakes include:

Adding a comma after the last key-value pair (JSON doesn't allow trailing commas).
Using "smart" or curly quotes “” instead of standard straight quotes "".

# Verify your JSON format
cat ~/.claude/settings.json | python3 -m json.tool

If the output looks normal, your format is fine. If it throws an error, you've got a syntax issue.

Q2: I get an Invalid API key error, but I'm sure my key is correct?

Possible reasons:

Extra spaces → Check if you accidentally copied a space at the beginning or end of the key.
Key expired or disabled → Log in to api.apiyi.com/token to regenerate it.
Environment variable priority → You might have an old ANTHROPIC_API_KEY set elsewhere in your system.

# Check for environment variables from multiple sources
env | grep ANTHROPIC

Q3: The model invocation works, but the results are poor or the formatting is weird?

This usually happens if the API proxy service doesn't fully support the Anthropic format, particularly regarding how system prompts are handled.

To verify, you can test the proxy service directly using curl:

curl https://api.apiyi.com/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: sk-your-key" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 100,
    "messages": [{"role": "user", "content": "Say hello"}]
  }'

A successful response should include a content field with the actual text output. If the response is an error or looks malformed, the issue lies with the proxy service.

🎯 Quick Tip: APIYI (apiyi.com) is fully compatible with the native Anthropic format, so the curl test above will work perfectly there. If you're testing other proxy services, use this command to quickly verify compatibility.

Q4: How do I set environment variables on Windows?

For Windows users, using the settings.json method is much simpler and more reliable:

// C:\Users\YourUsername\.claude\settings.json
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.apiyi.com",
    "ANTHROPIC_API_KEY": "sk-your-API-key"
  }
}

If you prefer using PowerShell to set temporary environment variables:

$env:ANTHROPIC_BASE_URL = "https://api.apiyi.com"
$env:ANTHROPIC_API_KEY = "sk-your-API-key"
claude

Q5: How can I use different API configurations for different projects?

Create a .claude/settings.json file in your project's root directory. This file will take priority over your global configuration:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.apiyi.com",
    "ANTHROPIC_API_KEY": "sk-project-specific-key"
  }
}

Note: If this file contains a key, make sure to add it to your .gitignore so you don't accidentally commit it to your repository. Using .claude/settings.local.json (for local-only config) is even safer, as it's ignored by git by default.

7. Configuration Verification Checklist

Once you've finished the configuration, follow these steps to verify everything is working:

# Step 1: Confirm environment variables are active
echo "Base URL: $ANTHROPIC_BASE_URL"
echo "API Key: ${ANTHROPIC_API_KEY:0:10}..."

# Step 2: Test API connectivity
curl -s https://api.apiyi.com/v1/models \
  -H "x-api-key: $ANTHROPIC_API_KEY" | head -c 200

# Step 3: Launch Claude Code and test
claude --version
claude

Type /status in the Claude Code prompt to check the current connection status and confirm that the Base URL and model configuration are correct.

🎯 Pro Tip: After configuration, we recommend sending a simple test message to ensure Claude Code is responding correctly. The APIYI (apiyi.com) platform supports balance queries, allowing you to monitor Claude Code's token usage in real-time via the dashboard to keep costs under control.

Summary

If you encounter a "model not found" error when connecting Claude Code to a third-party API proxy service, 90% of the time it's because the Base URL format is incorrect. Keep these core principles in mind:

Don't add /v1 to the Base URL: Use https://api.apiyi.com. Claude Code will automatically append /v1/messages.
Exact model name matching: Use full names like claude-sonnet-4-6, claude-opus-4-6, or claude-haiku-4-5-20251001.
Recommended settings.json configuration: Writing to ~/.claude/settings.json makes the settings permanent, so you don't have to set environment variables every time.
Dedicated Claude Code tokens: Select the ClaudeCode group in the APIYI dashboard to enjoy a 12% discount.

If you're using the official Anthropic API key and have a direct network connection, you can simply run the /model command within Claude Code to select your model without any extra configuration.

🎯 Get your API Key: Visit APIYI (apiyi.com) to register an account and create a ClaudeCode group token on the token management page (api.apiyi.com/token). The platform supports pay-as-you-go with no minimum spend, billing based on actual token usage—perfect for both individuals and teams.

This article was compiled by the APIYI technical team based on actual testing with Claude Code v2.x. If you run into any configuration issues, feel free to visit the APIYI Help Center at help.apiyi.com for support.

3 methods to solve model not found issues when configuring Claude Code with third-party API proxy services

I. Quick Diagnosis: Is it the Model Name or the Base URL?

2. Core Configuration for Connecting Third-Party Proxy APIs to Claude Code

2.1 Correct Base URL Format: Drop the /v1

2.2 Getting Your API Key

III. Detailed Guide to the Two Configuration Methods

Method 1: Environment Variable Configuration (Recommended for Temporary Testing)

Method 2: settings.json Configuration (Recommended for Long-term Use)

IV. Latest Model Names Supported by Claude Code

Standard Reasoning Models

Enhanced Reasoning Models (Thinking Mode)

5. Switching Models in Claude Code

5.1 Using the /model Command (The Easiest Way)

5.2 Specifying Default Models in settings.json

6. Common Errors and Solutions (FAQ)

7. Configuration Verification Checklist

Summary

3 steps to configure OpenCode to connect to API relay station and unlock free switching between 400+ AI models

5 Ways to Resolve the invalid beta flag Error When OpenClaw Calls the Claude API

Mastering the 5 Core Points of Claude Opus 4.6 Agent Teams Multi-agent Collaboration

Decoding ClawHub.ai: The official skill store of 220k-star OpenClaw, how 3000+ Skills expand your AI Agent

Unveiling Lovart AI Unlimited Plan “Word Games” – Nano Banana Pro Credit Limit Analysis and 3 Alternatives

Claude Opus 4.6 vs 4.5 Comprehensive Comparison: 12 Benchmarks Reveal the Real Gap

I. Quick Diagnosis: Is it the Model Name or the Base URL?

2. Core Configuration for Connecting Third-Party Proxy APIs to Claude Code

2.1 Correct Base URL Format: Drop the /v1

2.2 Getting Your API Key

III. Detailed Guide to the Two Configuration Methods

Method 1: Environment Variable Configuration (Recommended for Temporary Testing)

Method 2: settings.json Configuration (Recommended for Long-term Use)

IV. Latest Model Names Supported by Claude Code

Standard Reasoning Models

Enhanced Reasoning Models (Thinking Mode)

5. Switching Models in Claude Code

5.1 Using the /model Command (The Easiest Way)

5.2 Specifying Default Models in settings.json

6. Common Errors and Solutions (FAQ)

7. Configuration Verification Checklist

Summary

Similar Posts