chore: add claude 4 to verified mode & global replace 3.7 to claude 4 (#8665)

Co-authored-by: openhands <openhands@all-hands.dev>

.github/workflows/openhands-resolver.yml

@@ -24,7 +24,7 @@ on:
       LLM_MODEL:
         required: false
         type: string
-        default: "anthropic/claude-3-7-sonnet-20250219"
+        default: "anthropic/claude-sonnet-4-20250514"
       LLM_API_VERSION:
         required: false
         type: string

README.md

@@ -67,7 +67,7 @@ docker run -it --rm --pull=always \
 You'll find OpenHands running at [http://localhost:3000](http://localhost:3000)!
 
 When you open the application, you'll be asked to choose an LLM provider and add an API key.
-[Anthropic's Claude 3.7 Sonnet](https://www.anthropic.com/api) (`anthropic/claude-3-7-sonnet-20250219`)
+[Anthropic's Claude Sonnet 4](https://www.anthropic.com/api) (`anthropic/claude-sonnet-4-20250514`)
 works best, but you have [many options](https://docs.all-hands.dev/modules/usage/llms).
 
 ## 💡 Other ways to run OpenHands

docs/README.md

@@ -52,4 +52,4 @@ $ poetry run python docs/translation_updater.py
 # ...
 ```
 
-This process uses `claude-3-7-sonnet-20250219` as base model and each language consumes at least ~30k input tokens and ~35k output tokens.
+This process uses `claude-sonnet-4-20250514` as base model and each language consumes at least ~30k input tokens and ~35k output tokens.

docs/i18n/fr/docusaurus-plugin-content-docs/current/usage/llms/llms.md

@@ -13,7 +13,7 @@ recommandations pour la sélection de modèles. Nos derniers résultats d'évalu
 
 Sur la base de ces résultats et des retours de la communauté, les modèles suivants ont été vérifiés comme fonctionnant raisonnablement bien avec OpenHands :
 
-- [anthropic/claude-3-7-sonnet-20250219](https://www.anthropic.com/api) (recommandé)
+- [anthropic/claude-sonnet-4-20250514](https://www.anthropic.com/api) (recommandé)
 - [gemini/gemini-2.5-pro](https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/)
 - [deepseek/deepseek-chat](https://api-docs.deepseek.com/)
 - [openai/o3-mini](https://openai.com/index/openai-o3-mini/)

docs/i18n/ja/docusaurus-plugin-content-docs/current/usage/llms/llms.md

@@ -13,7 +13,7 @@ OpenHandsはLiteLLMでサポートされているあらゆるLLMに接続でき
 
 これらの調査結果とコミュニティからのフィードバックに基づき、以下のモデルはOpenHandsでうまく動作することが確認されています：
 
-- [anthropic/claude-3-7-sonnet-20250219](https://www.anthropic.com/api) (推奨)
+- [anthropic/claude-sonnet-4-20250514](https://www.anthropic.com/api) (推奨)
 - [gemini/gemini-2.5-pro](https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/)
 - [deepseek/deepseek-chat](https://api-docs.deepseek.com/)
 - [openai/o3-mini](https://openai.com/index/openai-o3-mini/)

docs/i18n/pt-BR/docusaurus-plugin-content-docs/current/usage/llms/llms.md

@@ -13,7 +13,7 @@ recomendações para seleção de modelos. Nossos resultados de benchmarking mai
 
 Com base nessas descobertas e feedback da comunidade, os seguintes modelos foram verificados e funcionam razoavelmente bem com o OpenHands:
 
-- [anthropic/claude-3-7-sonnet-20250219](https://www.anthropic.com/api) (recomendado)
+- [anthropic/claude-sonnet-4-20250514](https://www.anthropic.com/api) (recomendado)
 - [gemini/gemini-2.5-pro](https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/)
 - [deepseek/deepseek-chat](https://api-docs.deepseek.com/)
 - [openai/o3-mini](https://openai.com/index/openai-o3-mini/)

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/current/usage/llms/llms.md

@@ -12,7 +12,7 @@ OpenHands 可以连接到任何 LiteLLM 支持的 LLM。但是，它需要一个
 
 基于这些发现和社区反馈，以下模型已被验证可以与 OpenHands 合理地配合使用：
 
-- [anthropic/claude-3-7-sonnet-20250219](https://www.anthropic.com/api)（推荐）
+- [anthropic/claude-sonnet-4-20250514](https://www.anthropic.com/api)（推荐）
 - [gemini/gemini-2.5-pro](https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/)
 - [deepseek/deepseek-chat](https://api-docs.deepseek.com/)
 - [openai/o3-mini](https://openai.com/index/openai-o3-mini/)

docs/modules/usage/how-to/cli-mode.md

@@ -23,7 +23,7 @@ This command opens an interactive prompt where you can type tasks or commands an
 
 1. Set the following environment variables in your terminal:
    - `SANDBOX_VOLUMES` to specify the directory you want OpenHands to access ([See using SANDBOX_VOLUMES for more info](../runtimes/docker#using-sandbox_volumes))
-   - `LLM_MODEL` - the LLM model to use (e.g. `export LLM_MODEL="anthropic/claude-3-7-sonnet-20250219"`)
+   - `LLM_MODEL` - the LLM model to use (e.g. `export LLM_MODEL="anthropic/claude-sonnet-4-20250514"`)
    - `LLM_API_KEY` - your API key (e.g. `export LLM_API_KEY="sk_test_12345"`)
 
 2. Run the following command:

docs/modules/usage/how-to/headless-mode.md

@@ -23,7 +23,7 @@ To run OpenHands in Headless mode with Docker:
 
 1. Set the following environment variables in your terminal:
    - `SANDBOX_VOLUMES` to specify the directory you want OpenHands to access ([See using SANDBOX_VOLUMES for more info](../runtimes/docker#using-sandbox_volumes))
-   - `LLM_MODEL` - the LLM model to use (e.g. `export LLM_MODEL="anthropic/claude-3-7-sonnet-20250219"`)
+   - `LLM_MODEL` - the LLM model to use (e.g. `export LLM_MODEL="anthropic/claude-sonnet-4-20250514"`)
    - `LLM_API_KEY` - your API key (e.g. `export LLM_API_KEY="sk_test_12345"`)
 
 2. Run the following Docker command:

docs/modules/usage/llms/llms.md

@@ -13,7 +13,7 @@ recommendations for model selection. Our latest benchmarking results can be foun
 
 Based on these findings and community feedback, these are the latest models that have been verified to work reasonably well with OpenHands:
 
-- [anthropic/claude-3-7-sonnet-20250219](https://www.anthropic.com/api) (recommended)
+- [anthropic/claude-sonnet-4-20250514](https://www.anthropic.com/api) (recommended)
 - [openai/o4-mini](https://openai.com/index/introducing-o3-and-o4-mini/)
 - [gemini/gemini-2.5-pro](https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/)
 - [deepseek/deepseek-chat](https://api-docs.deepseek.com/)

docs/translation_updater.py

@@ -57,7 +57,7 @@ def translate_content(content, target_lang):
     system_prompt = f'You are a professional translator. Translate the following content into {target_lang}. Preserve all Markdown formatting, code blocks, and front matter. Keep any {{% jsx %}} tags and similar intact. Do not translate code examples, URLs, or technical terms.'
 
     message = client.messages.create(
-        model='claude-3-7-sonnet-20250219',
+        model='claude-sonnet-4-20250514',
         max_tokens=4096,
         temperature=0,
         system=system_prompt,

frontend/__tests__/routes/llm-settings.test.tsx

@@ -48,7 +48,7 @@ describe("Content", () => {
 
       await waitFor(() => {
         expect(provider).toHaveValue("Anthropic");
-        expect(model).toHaveValue("claude-3-7-sonnet-20250219");
+        expect(model).toHaveValue("claude-sonnet-4-20250514");
 
         expect(apiKey).toHaveValue("");
         expect(apiKey).toHaveProperty("placeholder", "");
@@ -135,7 +135,7 @@ describe("Content", () => {
       );
       const condensor = screen.getByTestId("enable-memory-condenser-switch");
 
-      expect(model).toHaveValue("anthropic/claude-3-7-sonnet-20250219");
+      expect(model).toHaveValue("anthropic/claude-sonnet-4-20250514");
       expect(baseUrl).toHaveValue("");
       expect(apiKey).toHaveValue("");
       expect(apiKey).toHaveProperty("placeholder", "");
@@ -542,7 +542,7 @@ describe("Form submission", () => {
 
     // select model
     await userEvent.click(model);
-    const modelOption = screen.getByText("claude-3-7-sonnet-20250219");
+    const modelOption = screen.getByText("claude-sonnet-4-20250514");
     await userEvent.click(modelOption);
 
     const submitButton = screen.getByTestId("submit-button");
@@ -550,7 +550,7 @@ describe("Form submission", () => {
 
     expect(saveSettingsSpy).toHaveBeenCalledWith(
       expect.objectContaining({
-        llm_model: "anthropic/claude-3-7-sonnet-20250219",
+        llm_model: "anthropic/claude-sonnet-4-20250514",
         llm_base_url: "",
         confirmation_mode: false,
       }),

frontend/__tests__/utils/extract-model-and-provider.test.ts

@@ -71,6 +71,18 @@ describe("extractModelAndProvider", () => {
       separator: "/",
     });
 
+    expect(extractModelAndProvider("claude-sonnet-4-20250514")).toEqual({
+      provider: "anthropic",
+      model: "claude-sonnet-4-20250514",
+      separator: "/",
+    });
+
+    expect(extractModelAndProvider("claude-opus-4-20250514")).toEqual({
+      provider: "anthropic",
+      model: "claude-opus-4-20250514",
+      separator: "/",
+    });
+
     expect(extractModelAndProvider("claude-3-haiku-20240307")).toEqual({
       provider: "anthropic",
       model: "claude-3-haiku-20240307",

frontend/src/mocks/handlers.ts

@@ -100,7 +100,7 @@ const openHandsHandlers = [
       "gpt-4o",
       "gpt-4o-mini",
       "anthropic/claude-3.5",
-      "anthropic/claude-3-7-sonnet-20250219",
+      "anthropic/claude-sonnet-4-20250514",
     ]),
   ),
 

frontend/src/routes/llm-settings.tsx

@@ -279,7 +279,7 @@ function LlmSettingsScreen() {
                 <ModelSelector
                   models={modelsAndProviders}
                   currentModel={
-                    settings.LLM_MODEL || "anthropic/claude-3-5-sonnet-20241022"
+                    settings.LLM_MODEL || "anthropic/claude-sonnet-4-20250514"
                   }
                   onChange={handleModelIsDirty}
                 />
@@ -342,9 +342,9 @@ function LlmSettingsScreen() {
                 name="llm-custom-model-input"
                 label={t(I18nKey.SETTINGS$CUSTOM_MODEL)}
                 defaultValue={
-                  settings.LLM_MODEL || "anthropic/claude-3-7-sonnet-20250219"
+                  settings.LLM_MODEL || "anthropic/claude-sonnet-4-20250514"
                 }
-                placeholder="anthropic/claude-3-7-sonnet-20250219"
+                placeholder="anthropic/claude-sonnet-4-20250514"
                 type="text"
                 className="w-[680px]"
                 onChange={handleCustomModelIsDirty}

frontend/src/services/settings.ts

@@ -3,7 +3,7 @@ import { Settings } from "#/types/settings";
 export const LATEST_SETTINGS_VERSION = 5;
 
 export const DEFAULT_SETTINGS: Settings = {
-  LLM_MODEL: "anthropic/claude-3-7-sonnet-20250219",
+  LLM_MODEL: "anthropic/claude-sonnet-4-20250514",
   LLM_BASE_URL: "",
   AGENT: "CodeActAgent",
   LANGUAGE: "en",

frontend/src/utils/verified-models.ts

@@ -6,6 +6,8 @@ export const VERIFIED_MODELS = [
   "o4-mini-2025-04-16",
   "claude-3-5-sonnet-20241022",
   "claude-3-7-sonnet-20250219",
+  "claude-sonnet-4-20250514",
+  "claude-opus-4-20250514",
   "deepseek-chat",
 ];
 
@@ -39,4 +41,6 @@ export const VERIFIED_ANTHROPIC_MODELS = [
   "claude-3-opus-20240229",
   "claude-3-sonnet-20240229",
   "claude-3-7-sonnet-20250219",
+  "claude-sonnet-4-20250514",
+  "claude-opus-4-20250514",
 ];

openhands/cli/utils.py

@@ -167,6 +167,8 @@ VERIFIED_ANTHROPIC_MODELS = [
     'claude-3-opus-20240229',
     'claude-3-sonnet-20240229',
     'claude-3-7-sonnet-20250219',
+    'claude-sonnet-4-20250514',
+    'claude-opus-4-20250514',
 ]
 
 

openhands/core/config/llm_config.py

@@ -47,7 +47,7 @@ class LLMConfig(BaseModel):
         seed: The seed to use for the LLM.
     """
 
-    model: str = Field(default='claude-3-7-sonnet-20250219')
+    model: str = Field(default='claude-sonnet-4-20250514')
     api_key: SecretStr | None = Field(default=None)
     base_url: str | None = Field(default=None)
     api_version: str | None = Field(default=None)

openhands/resolver/README.md

@@ -109,7 +109,7 @@ export GIT_USERNAME="your-gitlab-username"  # Optional, defaults to token owner
 
 # LLM configuration
 
-export LLM_MODEL="anthropic/claude-3-7-sonnet-20250219"  # Recommended
+export LLM_MODEL="anthropic/claude-sonnet-4-20250514"  # Recommended
 export LLM_API_KEY="your-llm-api-key"
 export LLM_BASE_URL="your-api-url"  # Optional, for API proxies
 ```

openhands/resolver/examples/openhands-resolver.yml

@@ -24,7 +24,7 @@ jobs:
       macro: ${{ vars.OPENHANDS_MACRO || '@openhands-agent' }}
       max_iterations: ${{ fromJson(vars.OPENHANDS_MAX_ITER || 50) }}
       base_container_image: ${{ vars.OPENHANDS_BASE_CONTAINER_IMAGE || '' }}
-      LLM_MODEL: ${{ vars.LLM_MODEL || 'anthropic/claude-3-7-sonnet-20250219' }}
+      LLM_MODEL: ${{ vars.LLM_MODEL || 'anthropic/claude-sonnet-4-20250514' }}
       target_branch: ${{ vars.TARGET_BRANCH || 'main' }}
       runner: ${{ vars.TARGET_RUNNER }}
     secrets:

tests/unit/test_cli_utils.py

@@ -354,11 +354,11 @@ class TestModelAndProviderFunctions:
         assert result['separator'] == '/'
 
     def test_extract_model_and_provider_anthropic_implicit(self):
-        model = 'claude-3-7-sonnet-20250219'
+        model = 'claude-sonnet-4-20250514'
         result = extract_model_and_provider(model)
 
         assert result['provider'] == 'anthropic'
-        assert result['model'] == 'claude-3-7-sonnet-20250219'
+        assert result['model'] == 'claude-sonnet-4-20250514'
         assert result['separator'] == '/'
 
     def test_extract_model_and_provider_versioned(self):
@@ -380,7 +380,7 @@ class TestModelAndProviderFunctions:
     def test_organize_models_and_providers(self):
         models = [
             'openai/gpt-4o',
-            'anthropic/claude-3-7-sonnet-20250219',
+            'anthropic/claude-sonnet-4-20250514',
             'o3-mini',
             'anthropic.claude-3-5',  # Should be ignored as it uses dot separator for anthropic
             'unknown-model',
@@ -397,7 +397,7 @@ class TestModelAndProviderFunctions:
         assert 'o3-mini' in result['openai']['models']
 
         assert len(result['anthropic']['models']) == 1
-        assert 'claude-3-7-sonnet-20250219' in result['anthropic']['models']
+        assert 'claude-sonnet-4-20250514' in result['anthropic']['models']
 
         assert len(result['other']['models']) == 1
         assert 'unknown-model' in result['other']['models']