The Token Ledger Digest – 2026-07-15

Most cost‑impacting change

• NVIDIA: Nemotron 3 Ultra – Completion price rose from $2.20 / 1M to $3.60 / 1M (+$1.40); prompt price rose from $0.50 / 1M to $0.60 / 1M (+$0.10). Who should care: Users running long‑form generation workloads on this model will see ~64% higher per‑token cost.

Other notable price moves

Model	Change	Old ($/1M)	New ($/1M)	Δ ($/1M)
Z.ai: GLM 5	Prompt ↑	0.60	0.95	+0.35
	Completion ↑	1.92	3.15	+1.23
MoonshotAI: Kimi K2.5	Prompt ↑	0.38	0.57	+0.19
	Completion ↑	2.03	2.85	+0.83
Qwen: Qwen3.5 397B A17B	Prompt ↑	0.39	0.45	+0.07
	Completion ↑	2.45	3.00	+0.55
Tencent: Hy3	Prompt ↑	0.14	0.20	+0.06
	Completion ↑	0.58	0.80	+0.22
MiniMax: MiniMax M2.7	Prompt ↑	0.24	0.30	+0.06
	Completion ↑	0.96	1.20	+0.24
Qwen: Qwen3.6 27B	Prompt ↑	0.29	0.45	+0.16
	Completion ↑	2.40	2.70	+0.30
Z.ai: GLM 5.2	Prompt ↓	0.93	0.89	–0.04
	Completion ↓	2.92	2.79	–0.13
Google: Gemma 4 26B A4B	Prompt ↑	0.06	0.10	+0.04
	Completion ↓	0.33	0.30	–0.03

Removed model

• Sao10K: Llama 3.1 70B Hanami x1 – Deleted; previously $3.00 / 1M for both prompt and completion. Who should care: Anyone relying on this specific 70B variant must migrate to an alternative.

Cheapest models today (per‑million‑token rates)

1. inclusionAI: Ling-2.6-flash – $0.01 prompt / $0.03 completion
2. IBM: Granite 4.0 Micro – $0.017 prompt / $0.112 completion
3. Mistral: Mistral Nemo – $0.02 prompt / $0.04 completion

Total models tracked: 343.

---

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