The Token Ledger Digest – 2026-07-16

Lead change – biggest cost impact:

Qwen: Qwen3.7 Max – Prompt rose from $1.25/M to $1.475/M (+$0.225/M); completion rose from $3.75/M to $4.425/M (+$0.675/M).

Who should care: Teams running high‑volume inference on this model will see ~9% higher per‑million‑token spend.

Other price changes

• MoonshotAI: Kimi K2.6 – Prompt $0.66/M → $0.95/M (+$0.29/M); Completion $3.41/M → $4.00/M (+$0.59/M).
  
  Who should care: Cost‑sensitive apps using long completions.

• Z.ai: GLM 5.2 – Prompt $0.8876/M → $0.959/M (+$0.0714/M); Completion $2.7896/M → $3.014/M (+$0.2244/M).
  
  Who should care: Moderate‑traffic workloads.

• Google: Gemma 4 31B – Prompt $0.12/M → $0.22/M (+$0.10/M); Completion $0.37/M → $0.55/M (+$0.18/M).
  
  Who should care: Users of this mid‑size model.

• Qwen: Qwen3 Coder Next – Prompt $0.12/M → $0.11/M (−$0.01/M); Completion unchanged $0.80/M.
  
  Who should care: Slight savings for prompt‑heavy tasks.

• Z.ai: GLM 4.7 Flash – Prompt $0.0605/M → $0.0600/M (−$0.0005/M); Completion unchanged $0.40/M.
  
  Who should care: Negligible impact.

• Qwen: Qwen3 14B – Prompt $0.10/M → $0.12/M (+$0.02/M); Completion unchanged $0.24/M.
  
  Who should care: Small uptick for prompt‑driven use.

• Meta: Llama 3.2 3B Instruct – Prompt $0.0509/M → $0.0500/M (−$0.0009/M); Completion $0.335/M → $0.330/M (−$0.005/M).
  
  Who should care: Minor cost reduction.

Removed model

Arcee AI: Coder Large – Prompt $0.50/M, Completion $0.80/M, 32k context.

Who should care: Anyone relying on this model must migrate to an alternative.

No other changes reported. Total models tracked: 342.

---

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