LiteLLM Fixes Token Bloat and Cost Tracking Across Claude, Azure OpenAI, Fireworks

Key Info

LiteLLM published a roundup of four fixes made in the last 24 hours covering token usage and cost tracking for Claude, Azure OpenAI and Fireworks models.

Highlights

• Codex sessions on Claude now start at 30.8K tokens, down from 688K (#48972)
• PDFs from Claude apps now work on Azure OpenAI (#47166)
• PDF token estimates on Azure GPT now match the bill: 141, down from 1664 (#47554)
• Fireworks models now show real spend instead of $0 (#48900)