* fix(deepseek): add reasoning token cost for deepseek-v4-pro
DeepSeek bills reasoning (CoT) tokens at the standard output rate.
https://api-docs.deepseek.com/quick_start/pricing/
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(deepseek): add reasoning token cost for deepseek-v4-flash
DeepSeek bills reasoning (CoT) tokens at the standard output rate.
https://api-docs.deepseek.com/quick_start/pricing/
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(deepseek): cover reasoner and cite reasoning=output billing
Add cost.reasoning for deepseek-reasoner (same gap as V4) and document
that CoT is billed at the output rate with reasoning_tokens as a
completion_tokens subset, so estimators do not double-count.
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Aiden Cline <aidenpcline@gmail.com>
- Updated release_date and last_updated fields for Anthropic Claude 2.1, Claude 2, DeepSeek Chat, Codestral, and Vercel models.
- Ensured all dates are accurate and follow the YYYY-MM-DD format.
- Updated 151+ model files across 11 providers with accurate release_date and last_updated fields
- Added dates for OpenAI, Azure, Anthropic, Google, Amazon Bedrock, Meta Llama, DeepSeek, xAI, Mistral, Groq, and other providers
- Corrected Llama 4 model dates to accurate release date (2025-04-05)
- Aligned Azure OpenAI models with corresponding OpenAI model dates
- Verified DeepSeek model dates through web research
- All dates follow YYYY-MM-DD format as per schema requirements
- Rename inputCached/outputCached to cache_read/cache_write in schema
- Make cache fields optional in ModelSchema
- Remove cache fields from TOML files where value is $0
- Keep cache fields only when they have actual pricing values
🤖 Generated with [opencode](https://opencode.ai)
Co-Authored-By: opencode <noreply@opencode.ai>
Added npm field to ProviderSchema validation and specified AI SDK packages for all providers to enable package dependency tracking.
🤖 Generated with [opencode](https://opencode.ai)
Co-Authored-By: opencode <noreply@opencode.ai>
Xarth