The Token Ledger Digest – 2026-08-02

Price Drop – Z.ai: GLM 5.2

• Prompt: fell from $0.76 /1M to $0.28 /1M (‑63%).
• Completion: fell from $2.39 /1M to $0.89 /1M (‑63%).
• Who should care: Teams running large‑scale inference or fine‑tuning on GLM 5.2 see per‑token costs cut by roughly two‑thirds.

Price Drop – DeepSeek: DeepSeek V4 Flash 0731

• Prompt: dropped from $0.14 /1M to $0.09 /1M (‑36%).
• Completion: dropped from $0.28 /1M to $0.18 /1M (‑36%).
• Who should care: Users of the 0731 checkpoint benefit from lower latency‑cost trade‑offs for flash‑style workloads.

Price Increase – NVIDIA: Nemotron 3 Ultra

• Prompt: rose from $0.50 /1M to $0.60 /1M (+20%).
• Completion: rose from $2.20 /1M to $3.60 /1M (+64%).
• Who should care: Cost‑sensitive applications relying on Nemotron 3 Ultra for completion‑heavy tasks will see noticeably higher bills.

Added Model – DeepSeek V4 Flash Latest

• Prompt: $0.09 /1M
• Completion: $0.18 /1M
• Context: 1,048,576 tokens
• Who should care: Developers needing ultra‑long context with flash‑level pricing can now access this newest variant.

Total models tracked: 337.

---

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