🎯 API Improvement Request - Token Factory
Summary: The Token Factory /models endpoint orders models in a way that buries popular and important models deep in the response (positions 20-25), causing developer confusion and integration issues.
Technical Report: https://gist.github.com/opencolin/a1ce112f0bc3459c59ff22f618c54a8c
Current Problem:
Popular models appear at the END of the 25-model response:
- Position 20:
moonshotai/Kimi-K2.6 (popular general model)
- Position 23:
zai-org/GLM-5.2 (latest GLM model)
- Position 24:
moonshotai/Kimi-K2.7-Code (popular coding model)
- Position 25:
deepseek-ai/DeepSeek-V4-Pro (latest DeepSeek)
Real Developer Impact:
- 🤖 AI Agents: Conclude models "don't exist" when using
.slice(0,15) or head -20
- 👨💻 Developers: Build UIs showing less useful models first
- 📱 Integrations: Popular models are not discoverable in typical pagination
- 🔍 Discovery: Key flagship models are hidden from casual browsing
Simple Solution:
Reorder models by popularity/strategic importance:
deepseek-ai/DeepSeek-V4-Prozai-org/GLM-5.2moonshotai/Kimi-K2.7-Codemeta-llama/Llama-3.3-70B-Instructmoonshotai/Kimi-K2.6- ... (remaining models)
Industry Comparison:
- OpenAI: Shows
gpt-4, gpt-3.5-turbo first (most popular models)
- Anthropic: Shows
claude-3-5-sonnet first (flagship model)
- Token Factory: Popular models buried at positions 20-25 😞
Business Benefits:
- ✅ Better developer experience and model discoverability
- ✅ Reduced integration confusion and support tickets
- ✅ Improved competitive positioning vs OpenAI/Anthropic APIs
- ✅ Better AI agent compatibility (prevents truncation errors)
Implementation:
Simple reordering of existing data - high impact, low effort improvement.
Priority: Medium-High
Effort: Low (backend ordering change)
Impact: High (affects all Token Factory integrations)
This issue was discovered during AI agent integration work where truncated API responses caused incorrect "model not available" conclusions for flagship models.
🎯 API Improvement Request - Token Factory
Summary: The Token Factory
/modelsendpoint orders models in a way that buries popular and important models deep in the response (positions 20-25), causing developer confusion and integration issues.Technical Report: https://gist.github.com/opencolin/a1ce112f0bc3459c59ff22f618c54a8c
Current Problem:
Popular models appear at the END of the 25-model response:
moonshotai/Kimi-K2.6(popular general model)zai-org/GLM-5.2(latest GLM model)moonshotai/Kimi-K2.7-Code(popular coding model)deepseek-ai/DeepSeek-V4-Pro(latest DeepSeek)Real Developer Impact:
.slice(0,15)orhead -20Simple Solution:
Reorder models by popularity/strategic importance:
deepseek-ai/DeepSeek-V4-Prozai-org/GLM-5.2moonshotai/Kimi-K2.7-Codemeta-llama/Llama-3.3-70B-Instructmoonshotai/Kimi-K2.6Industry Comparison:
gpt-4,gpt-3.5-turbofirst (most popular models)claude-3-5-sonnetfirst (flagship model)Business Benefits:
Implementation:
Simple reordering of existing data - high impact, low effort improvement.
Priority: Medium-High
Effort: Low (backend ordering change)
Impact: High (affects all Token Factory integrations)
This issue was discovered during AI agent integration work where truncated API responses caused incorrect "model not available" conclusions for flagship models.