Scale computer-use 2.0 with open-source drivers, cross-OS sandboxes, and benchmarks for training, evaluation, and data generation.
Rust agentai-agentapplecomputer-usecomputer-use-agentcontainerizationcuadesktop-automationhacktoberfestlumemacosmanusoperatorswiftvirtualizationvirtualization-frameworkwindowswindows-sandbox
Update model list and schedule daily test
Adam762daaa99ae8
diff --git a/.github/workflows/test-cua-models.yml b/.github/workflows/test-cua-models.ymlindex 2fa3f2067b..0b5f447bea 100644--- a/.github/workflows/test-cua-models.yml+++ b/.github/workflows/test-cua-models.yml@@ -13,39 +13,44 @@ on:required: falsedefault: truetype: boolean+ schedule:+ # Runs at 4 PM UTC (8 AM PST) from November to March+ - cron: "0 16 * * *"+ # Runs at 3 PM UTC (8 AM PDT) from March to November+ - cron: "0 15 * * *"jobs:- # Test all CUA models - runs on PRs or when manually triggered+ # Test all CUA models - runs on PRs, schedules, or when manually triggeredtest-all-models:- if: ${{ github.event_name == 'pull_request_target' || fromJSON(inputs.test_models || 'false') }}+ if: ${{ github.event_name == 'pull_request_target' || github.event_name == 'schedule' || fromJSON(inputs.test_models || 'false') }}runs-on: ubuntu-lateststrategy:fail-fast: falsematrix:model:# Claude Sonnet/Haiku- # - anthropic/claude-sonnet-4-5-20250929+ - anthropic/claude-sonnet-4-5-20250929- anthropic/claude-haiku-4-5-20251001- # - anthropic/claude-opus-4-1-20250805+ - anthropic/claude-opus-4-1-20250805# OpenAI CU Preview- openai/computer-use-preview# GLM-V- # - openrouter/z-ai/glm-4.5v+ - openrouter/z-ai/glm-4.5v# - huggingface-local/zai-org/GLM-4.5V # Requires local model setup# Gemini CU Preview# - gemini-2.5-computer-use-preview-10-2025# InternVL- # - huggingface-local/OpenGVLab/InternVL3_5-1B+ - huggingface-local/OpenGVLab/InternVL3_5-1B# - huggingface-local/OpenGVLab/InternVL3_5-2B# - huggingface-local/OpenGVLab/InternVL3_5-4B# - huggingface-local/OpenGVLab/InternVL3_5-8B# UI-TARS (supports full computer-use, can run standalone)- # - huggingface-local/ByteDance-Seed/UI-TARS-1.5-7B+ - huggingface-local/ByteDance-Seed/UI-TARS-1.5-7B# Note: OpenCUA, GTA, and Holo are grounding-only models# They only support predict_click(), not agent.run()@@ -53,21 +58,21 @@ jobs:# Moondream (typically used in composed agents)# Format: moondream3+{any-llm-with-tools}- # - moondream3+anthropic/claude-sonnet-4-5-20250929 # Claude has VLM + Tools+ - moondream3+anthropic/claude-sonnet-4-5-20250929 # Claude has VLM + Tools# - moondream3+openai/gpt-4o # GPT-4o has VLM + Tools# OmniParser (typically used in composed agents)# Format: omniparser+{any-vlm-with-tools}- # - omniparser+anthropic/claude-sonnet-4-5-20250929 # Claude has VLM + Tools+ - omniparser+anthropic/claude-sonnet-4-5-20250929 # Claude has VLM + Tools# - omniparser+openai/gpt-4o # GPT-4o has VLM + Tools# Other grounding models + VLM with tools# Format: {grounding-model}+{any-vlm-with-tools}# These grounding-only models (OpenCUA, GTA, Holo) must be used in composed form# since they only support predict_click(), not full agent.run()- # - huggingface-local/HelloKKMe/GTA1-7B+anthropic/claude-sonnet-4-5-20250929- # - huggingface-local/xlangai/OpenCUA-7B+anthropic/claude-sonnet-4-5-20250929- # - huggingface-local/Hcompany/Holo1.5-3B+anthropic/claude-sonnet-4-5-20250929+ - huggingface-local/HelloKKMe/GTA1-7B+anthropic/claude-sonnet-4-5-20250929+ - huggingface-local/xlangai/OpenCUA-7B+anthropic/claude-sonnet-4-5-20250929+ - huggingface-local/Hcompany/Holo1.5-3B+anthropic/claude-sonnet-4-5-20250929steps:- name: Checkout repository@@ -109,7 +114,7 @@ jobs:uv venv --python 3.12uv pip install -e libs/python/agent -e libs/python/computeruv pip install -e libs/python/core- uv pip install "cua-agent[uitars-hf]"+ uv pip install "cua-agent[uitars-hf,internvl-hf,opencua-hf,moondream3,omni]"uv pip install pytest- name: Cache HuggingFace models@@ -139,6 +144,7 @@ jobs:OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}+ HF_TOKEN: ${{ secrets.HF_TOKEN }}- name: Calculate test duration and prepare messageif: always()@@ -244,7 +250,7 @@ jobs:# Summary job that aggregates all model test resultstest-summary:- if: ${{ always() && (github.event_name == 'pull_request_target' || fromJSON(inputs.test_models || 'false')) }}+ if: ${{ always() && (github.event_name == 'pull_request_target' || github.event_name == 'schedule' || fromJSON(inputs.test_models || 'false')) }}needs: test-all-modelsruns-on: ubuntu-lateststeps: