The Token Ledger – 2026-07-17

Most cost‑impacting change

• MoonshotAI Kimi Latest (~moonshotai/kimi-latest) – Prompt price rose from $0.66 to $3.00 per 1M tokens (+$2.34); completion price rose from $3.41 to $15.00 per 1M tokens (+$11.59). Who should care: Anyone using this model for long‑form generation; per‑million cost jumps ~14×.

Added models

• MoonshotAI: Kimi K3 (moonshotai/kimi-k3) – Prompt $3.00/1M, Completion $15.00/1M, 1M‑token context.
• Meta: Muse Spark 1.1 (meta/muse-spark-1.1) – Prompt $1.25/1M, Completion $4.25/1M, 1M‑token context. Who should care: Teams needing large‑context models; compare pricing against existing 1M‑context options.

Price changes (per‑million token values)

• Z.ai: GLM 5.2 (z-ai/glm-5.2) – Prompt $0.96 → $0.96 (−$0.001); Completion $3.01 → $3.01 (−$0.004).
• MoonshotAI: Kimi K2.7 Code (moonshotai/kimi-k2.7-code) – Prompt $0.72 → $0.75 (+$0.03); Completion $3.49 → $3.50 (+$0.01).
• MiniMax: MiniMax M2.7 (minimax/minimax-m2.7) – Prompt $0.30 → $0.25 (−$0.05); Completion $1.20 → $1.00 (−$0.20).
• Qwen: Qwen3.5 397B A17B (qwen/qwen3.5-397b-a17b) – Prompt $0.45 → $0.39 (−$0.06); Completion $3.00 → $2.34 (−$0.66).
• Qwen: Qwen3 14B (qwen/qwen3-14b) – Prompt $0.12 → $0.10 (−$0.02); Completion unchanged $0.24.
• Google: Gemma 3 27B (google/gemma-3-27b-it) – Prompt $0.08 → $0.10 (+$0.02); Completion $0.45 → $0.30 (−$0.15).
• Meta: Llama 3.2 3B Instruct (meta-llama/llama-3.2-3b-instruct) – Prompt $0.05 → $0.051 (+$0.001); Completion $0.33 → $0.335 (+$0.005).
• Mistral: Mistral Nemo (mistralai/mistral-nemo) – Prompt $0.02 → $0.019 (−$0.001); Completion $0.04 → $0.03 (−$0.01).

Who should care: Developers monitoring cost trends; note the modest decreases for several efficient models (MiniMax, Qwen families) and slight increases for others. No other meaningful changes reported.

---

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