Skip to content

Update 302ai model metadata and add GPT-5.4 configs#1497

Open
Lydanne wants to merge 3 commits intoanomalyco:devfrom
Lydanne:fix/302ai-models
Open

Update 302ai model metadata and add GPT-5.4 configs#1497
Lydanne wants to merge 3 commits intoanomalyco:devfrom
Lydanne:fix/302ai-models

Conversation

@Lydanne
Copy link
Copy Markdown
Contributor

@Lydanne Lydanne commented Apr 20, 2026

Summary

This PR updates the 302ai model catalog to match the latest provider capabilities and metadata.

Changes

  • add gpt-5.4 and gpt-5.4-pro model configs
  • add family metadata to Claude, GLM, and GPT model definitions
  • update capability flags such as reasoning, structured_output, temperature, and open_weights
  • normalize knowledge values to more precise dates
  • adjust token limits for context and input sizes
  • expand supported modalities for models with pdf or video input
  • add interleaved reasoning metadata for supported GLM models

Notes

  • Claude models now reflect updated reasoning support, knowledge cutoffs, and context windows
  • GLM models now reflect open-weight availability, updated limits, and multimodal support
  • GPT model configs now reflect reasoning-oriented behavior and current structured output support

Validation

  • Pass

Lydanne added 3 commits April 20, 2026 15:48
- Enable reasoning capability for GLM-4.5-air, GLM-4.5, GLM-4.5V, and GLM-4.6V models
- Update knowledge cutoff to 2025-04 for GLM-4.5, GLM-4.5V, GLM-4.6, GLM-4.6V, and GLM-4.7
- Add video input modality support for GLM-4.5V and GLM-4.6V
- Add structured_output capability for GLM-5-turbo and GLM-5.1
- Add interleaved reasoning_content field for GLM-4.7, GLM-5, GLM-5-turbo, GLM-5.1, and GLM-5V-turbo
- Add structured_output capability for GPT-4.1, GPT-4o, and GPT-5 series models
- Enable reasoning capability and disable temperature for GPT-5 series models
- Update context limits: GPT-4.1 series to 1,047,576 tokens, GPT-5.4 series to 1,050,000 tokens
- Add input token limits for GPT-5 series models (272,000 or 922,000 tokens)
- Update knowledge cutoffs across GPT-5 series (2024-05-30 to 2025-08-31)
- Add PDF input modality support for GPT-4
- Add family field for all Claude models (claude-haiku, claude-opus, claude-sonnet)
- Standardize knowledge cutoff dates to full date format (YYYY-MM-DD)
- Enable reasoning capability for Claude Opus 4.x and Sonnet 4.x series models
- Add PDF input modality support for claude-opus-4-1-20250805
- Update claude-opus-4-7 context limit to 1,000,000 tokens
@Lydanne
Copy link
Copy Markdown
Contributor Author

Lydanne commented Apr 20, 2026

@rekram1-node Please review and merge.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant