The Token Ledger Digest – 2026-07-19

The most cost‑impacting change is a price reduction for MoonshotAI’s Kimi K2.6 model.

• MoonshotAI: Kimi K2.6 – Prompt ↓ $0.95 → $0.684 /1M; Completion ↓ $4.00 → $3.42 /1M.
  
  Who should care: Teams running large‑scale code or reasoning workloads seeking lower per‑token spend.

• MoonshotAI: Kimi K2.7 Code – Prompt ↓ $1.00 → $0.85 /1M; Completion ↓ $4.40 → $3.80 /1M.
  
  Who should care: Developers using the model for coding assistants or API‑heavy applications.

• Z.ai: GLM 5.2 – Prompt ↓ $0.3374 → $0.2716 /1M; Completion ↓ $1.0604 → $0.8536 /1M.
  
  Who should care: Users of multilingual models looking to cut inference costs.

• Google: Gemma 4 26B A4B – Prompt ↓ $0.10 → $0.07 /1M; Completion ↑ $0.30 → $0.34 /1M.
  
  Who should care: Applications where prompt usage dominates; completion cost rise is modest.

• Google: Gemma 4 31B – Prompt ↑ $0.12 → $0.22 /1M; Completion ↑ $0.37 → $0.55 /1M.
  
  Who should care: Budget‑sensitive projects may need to reassess usage of this larger variant.

• MiniMax: MiniMax M2 – Prompt ↑ $0.255 → $0.30 /1M; Completion ↑ $1.02 → $1.20 /1M.
  
  Who should care: Enterprises relying on this model for high‑volume completions should monitor cost impact.

Originally published at The Token Ledger. Subscribe for the daily digest.