## Posts tagged #llm

1. [Jul 27, 2026  
   **Kimi K3 took first place on frontend coding at a third of the price, and the licence has a catch**  
   Kimi K3 took first place in the Frontend Code Arena ahead of Claude Fable 5 and GPT-5.6 Sol, at $3 and $15 per million tokens. The weights are out, and the licence is not open source.

2. [Jul 25, 2026  
   **LLM API pricing in 2026, compared across Claude, GPT, Gemini, and Kimi**  
   Verified per-million-token rates for Claude, GPT, and Gemini as of July 2026, which models actually compete with each other, and the billing rules that move bills more than any headline rate.

3. [Jul 25, 2026  
   **Claude Opus 5's effort levels, and what they actually cost you**  
   Claude Opus 5 costs $5/$25 per million tokens, the same as Opus 4.8. The effort parameter has five levels, not three, and changing it mid-conversation invalidates your prompt cache.

4. [Jul 24, 2026  
   **How to Fine-Tune an Open-Weights LLM: A Hands-On Guide Using Inkling**  
   A hands-on guide to fine-tuning an open-weights LLM with Thinking Machines' Inkling. Includes a fine-tune vs RAG vs prompt decision tree and the ROI math most guides skip.

5. [Jul 24, 2026  
   **Gemini API Pricing in 2026: What the New 3.6 Flash Rates Mean for Your Build Costs**  
   Google's Gemini 3.6 Flash dropped output pricing to $7.50/M and cut output token usage 17%. A real cost-per-workload breakdown vs GPT-5.6, with the math shown.