The Token Ledger – 2026-09-06

Qwen: Qwen3.5-35B-A3B – Prompt price rose from $0.08 to $0.3125 per 1M tokens (+$0.2325); completion price rose from $0.75 to $1.25 per 1M tokens (+$0.50). Impact: Developers using this model for generation‑heavy workloads will see per‑token costs increase by ~67% for prompts and ~67% for completions.

DeepSeek: DeepSeek V4 Pro 0423 – Prompt price fell from $0.8904 to $0.6876 per 1M tokens (-$0.2027); completion price fell from $1.7807 to $1.3753 per 1M tokens (-$0.4054). Impact: Cost‑sensitive teams relying on this model for long‑form completion can cut expenses by ~23%.

Z.ai: GLM Latest – Prompt price increased from $1.15 to $1.17 per 1M tokens (+$0.02); completion price increased from $3.50 to $3.96 per 1M tokens (+$0.46). Impact: Moderate uplift for completion‑focused use cases (~13% rise).

DeepSeek: DeepSeek V4 Flash 0731 – Prompt price dropped from $0.065 to $0.04998 per 1M tokens (-$0.0150); completion price dropped from $0.18 to $0.09996 per 1M tokens (-$0.0800). Impact: Notable savings (~44% completion reduction) for flash‑mode workloads.

DeepSeek V4 Flash Latest – Prompt price decreased from $0.04998 to $0.045 per 1M tokens (-$0.0050); completion price decreased from $0.09996 to $0.09 per 1M tokens (-$0.0100). Impact: Minor cost trim (~10% completion).

DeepSeek: DeepSeek V4 Flash 0423 – Prompt price slipped from $0.08526 to $0.08078 per 1M tokens (-$0.0045); completion price slipped from $0.17052 to $0.16156 per 1M tokens (-$0.0090). Impact: Small reduction (~5% completion).

Cheapest models today (per 1M tokens):

• IBM Granite 4.0 Micro – prompt $0.017, completion $0.112
• Mistral Nemo – prompt $0.019, completion $0.030
• inclusionAI Ling 3.0 Flash – prompt $0.021, completion $0.063

No models were added or removed; total model count remains 431.

---

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