A developer building a Rust-based video editing app (Kiru) shares how LLM token costs during development and testing were reaching ~$20/day. After trying local models (GLM, Qwen) and hitting concurrency limitations, they discovered Fireworks.ai's Firepass plan — an early-access $7/week unlimited token subscription for personal use. The plan supports OpenAI and Anthropic API schemas, works with coding agents like Claude Code and Codex, and is ideal for development/testing workflows. The author notes it's an early-access offering unlikely to remain in its current form indefinitely.

8m watch time

Sort: