The Token Ledger – 2026-08-07

Most cost‑impacting change: NVIDIA Nemotron 3 Super saw a price increase. Prompt rose from $0.085/1M to $0.30/1M (+$0.215/1M); completion rose from $0.40/1M to $0.90/1M (+$0.50/1M). Teams using this model for high‑volume inference should budget for roughly a 70% cost jump.

Other price changes

• MoonshotAI Kimi Latest: Prompt dropped from $2.90/1M to $2.50/1M (−$0.40/1M). Completion unchanged at $14.00/1M. Cost‑sensitive users of this model save ~14% on prompt tokens.
• Qwen VL 235B Thinking: Prompt fell from $0.98/1M to $0.40/1M (−$0.58/1M); completion rose slightly from $3.95/1M to $4.00/1M (+$0.05/1M). Net −$0.53/1M, benefiting vision‑language workloads.
• Thinking Machines Inkling Small: Prompt decreased from $0.50/1M to $0.45/1M (−$0.005/1M); completion steady at $1.20/1M. Minor saving for low‑traffic use.
• InclusionAI Ling‑3.0‑flash: Prompt cut from $0.075/1M to $0.021/1M (−$0.054/1M); completion cut from $0.22/1M to $0.063/1M (−$0.157/1M). Total −$0.211/1M – a notable reduction for existing users.
• Z.ai GLM 5.2: Prompt lowered from $0.76/1M to $0.679/1M (−$0.081/1M); completion lowered from $2.42/1M to $2.134/1M (−$0.286/1M). Net −$0.367/1M.
• Qwen 3.5‑122B: Prompt rose from $0.26/1M to $0.29/1M (+$0.03/1M); completion rose from $2.08/1M to $2.40/1M (+$0.32/1M). Net +$0.35/1M – a modest increase for those relying on this model.

Catalog updates: 61 new models were added, highlighted by the free inclusionai/ling-3.0-tiny, batch variants of Claude Opus 5, Gemini 3.6 Flash, and several OpenAI GPT‑5.6 Luna/Terra/Sol editions. One model was removed: the free inclusionai/ling-3.0-flash.

Developers should review the new batch offerings for potential latency‑cost tradeoffs and adjust usage of the adjusted models accordingly.

---

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