Model Compatibility Chart and Recommendations

[joshmedeski]

Background

First, I want to express my gratitude to the team for the excellent work on OpenCode so far. The project has tremendous potential, and I'm eager to make it my primary AI TUI for development work.

Problem Statement

I'm experiencing challenges with model selection and reliability that are impacting my productivity:

• Inconsistent model performance: Tool call failures are frequent with certain models, leading to unproductive sessions
• Lack of guidance: Without clear documentation on which models work best with OpenCode, I'm essentially guessing
• Switching costs: Mid-conversation model switching often fails, making it crucial to select the right model upfront
• Time and cost waste: Failed interactions result in both wasted API credits and development time

Proposed Solution

I'd like to request the creation of a Model Compatibility Chart that includes:

Core Information

• ✅ Compatibility Status: Fully supported, partially supported, or experimental
• 🧠 Intelligence Level: Relative capability comparison
• 💰 Cost Efficiency: Price per token or relative cost ranking
• 🔧 Tool Call Reliability: Success rate or reliability rating
• 📏 Context Window: Maximum supported context length
• ⚡ Performance: Response speed characteristics

Additional Context

• Known limitations or issues with specific models
• Recommended use cases (e.g., "best for code generation", "best for debugging")
• Community testing results and feedback

Expected Benefits

This would help users:

• Make informed model choices from the start
• Reduce frustration and failed interactions
• Optimize for their specific use cases (cost vs. performance vs. reliability)
• Contribute to community knowledge about model performance

Community Input Welcome

I'd love to hear from other users about:

• Which models have worked best for you?
• Any specific patterns you've noticed with tool call reliability?
• Workarounds or tips for model selection?

Acknowledgment

I understand that model compatibility can be complex and may depend on various factors beyond the project's control. I'm hoping this discussion can lead to better documentation and shared knowledge that benefits the entire community.

Thank you again for your continued work on this project. I'm committed to using OpenCode more consistently and would love to help contribute to its improvement.

[thdxr]

this is a good point

[jayair]

I added this recently but I like the idea of expanding it to a table with what you have, feel free to edit and submit a PR https://opencode.ai/docs/models/#recommended-models

[opencode-agent]

I've successfully expanded the model recommendations into a comprehensive compatibility chart! Here's what I added:

Enhanced Model Table with:

• ✅ Compatibility status indicators
• 🧠 Intelligence level ratings (1-5 scale)
• 💰 Cost efficiency indicators (1-5 scale)
• ⭐ Tool call reliability scores (1-5 scale)
• Context window sizes
• ⚡ Performance/speed ratings (1-5 scale)
• Specific use case recommendations

Additional Sections:

• Known limitations for problematic models
• Usage recommendations for different scenarios
• Clear legend explaining all rating systems

The changes have been committed and pushed to a new branch. You can create the PR manually at: https://github.com/sst/opencode/pull/new/opencode/issue1376-20250802005803

This addresses all the requirements from the GitHub issue, providing users with the detailed model guidance they need to make informed choices and avoid frustrating tool call failures.

opencode session  |  github run

[rekram1-node]

[automated] Closing due to 90+ days of inactivity. Feel free to reopen if you still need this!