Token Ledger Digest – 2026-08-15

Most cost‑impacting change:

• Z.ai: GLM 5.1 – Prompt price fell from $1.40 / 1M to $0.966 / 1M (−$0.434); completion price fell from $4.40 / 1M to $3.036 / 1M (−$1.364). Total cost ↓ $1.798 / 1M. Who should care: Teams running high‑volume inference on GLM 5.1 will see ~40% lower per‑million‑token spend.

Added Models

• Qwen: Qwen3.8 27B – Prompt $0.45 / 1M, Completion $3.20 / 1M, context 262,144 tokens.
• Dots Studio: Dots3‑Note Preview (free) – Prompt $0.00, Completion $0.00, context 512,000 tokens. Who should care: Developers needing a free, long‑context model for note‑taking or prototyping.

Price Changes (selected highlights)

Model	Old Prompt ($/1M)	New Prompt ($/1M)	Δ Prompt	Old Comp ($/1M)	New Comp ($/1M)	Δ Comp	Δ Total ($/1M)	Who should care
DeepSeek V4 Flash Latest	0.08	0.07	−0.01	0.252	0.14	−0.112	−0.122	Cost‑sensitive batch jobs
Z.ai: GLM 5.2	0.63	0.49	−0.14	1.98	1.54	−0.44	−0.58	Mid‑scale LLM users
MoonshotAI: Kimi K2.7 Code	0.67	0.71	+0.04	3.40	3.50	+0.10	+0.14	Teams needing slightly higher quality
DeepSeek V4 Flash 0423	0.14	0.074	−0.066	0.28	0.148	−0.132	−0.198	Flash inference workloads
Tencent: Hy3 preview	0.063	0.18	+0.117	0.21	0.60	+0.39	+0.507	Experimental preview users
MoonshotAI: Kimi K2.6	0.95	0.65	−0.30	4.00	3.41	−0.59	−0.89	Code‑gen pipelines
Qwen: Qwen3.5‑35B‑A3B	0.25	0.225	−0.025	1.25	1.80	+0.55	+0.525	Balanced prompt/completion shift
Qwen: Qwen3.5 397B A17B	0.50	0.39	−0.11	3.60	2.34	−1.26	−1.37	Large‑model cost optimizers
Z.ai: GLM 5	0.95	0.60	−0.35	2.55	1.92	−0.63	−0.98	General purpose GLM users
Z.ai: GLM 4.6	0.50	0.55	+0.05	2.00	2.20	+0.20	+0.25	Slight uptick for legacy GLM
Meta: Llama 4 Maverick	0.20	0.20	0.00	0.696	0.80	+0.104	+0.104	Maverick adopters

No models were removed today.

Total models tracked: 413.

---

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