Developer builds Khidi to make cheap open-weight AI models as easy to use as OpenAI
Open-weight language models have improved significantly in 2025-26, matching flagship closed models on several benchmarks at a fraction of the cost, yet most developers continue paying premium prices out of convenience. The main barrier is not model quality but the friction involved in setting up inference providers, managing API keys, and handling billing. A developer built Khidi (khidi.ai), an OpenAI-compatible API gateway that routes requests to low-cost open models like Kimi K2.6, requiring only a base URL change in existing code. The service claims to reduce a typical coding-agent monthly bill from around $500-550 with OpenAI or Anthropic to roughly $45 using open alternatives. The founder acknowledges that the latest closed frontier models still lead on the hardest tasks, and some provider-specific features may not translate perfectly through the gateway.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in