feat(together-ai): update model YAMLs [bot]

[hganwani-droid]

Auto-generated by poc-agent for provider together-ai.

---

Note

Medium Risk
Updates Together AI model metadata (pricing, context limits, features/modalities, and message options) and marks many models as deprecated, which can affect model selection and cost calculations but doesn’t change runtime code.

Overview
Refreshes Together AI model YAMLs to align with current provider metadata.

Adds/normalizes modalities and, for some models, messages.options, updates several limits (notably context windows/token caps) and costs fields, and adjusts advertised features (e.g., tool/function calling support).

Marks a large set of older/preview models as isDeprecated: true to discourage selection going forward.

^{Written by Cursor Bugbot for commit aae8396. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Image model missing input modality definition entirely

High Severity

The modalities section in imagen-4.0-preview.yaml had its input: key replaced with output: instead of keeping both. The model now has no input modalities defined at all. Every other image-mode model in this provider (imagen-4.0-fast, imagen-4.0-ultra, flash-image-2.5) defines input modalities. This model likely needs input: - text alongside the output: - image, matching the pattern in imagen-4.0-fast.yaml.

 

[cursor]

DeepSeek-R1 context window off-by-one error

Medium Severity

The context_window for DeepSeek-R1 was changed from 163840 to 163839, which is an off-by-one from the correct value. Together AI's documentation specifies 164K tokens (163840 = 160 × 1024), and every other DeepSeek model across all providers in this repository (aws-bedrock, deepinfra) consistently uses 163840. This is the only instance of 163839 in the entire codebase.