The Token Ledger – 2026-09-04
Most cost‑impacting change: MoonshotAI Kimi Latest completion price rose from $12.75 to $14.00 per 1M tokens (+$1.25/1M), while prompt price fell slightly from $2.55 to $2.50 per 1M (-$0.05/1M). Developers using long‑form completions should budget for higher output costs.
Price changes (per 1M tokens)
| Model | Prompt (old → new) | Completion (old → new) | Net impact |
|---|---|---|---|
| Z.ai: GLM Latest | $1.12 → $1.15 (+$0.03) | $3.52 → $3.50 (-$0.02) | +$0.01 |
| Qwen: Qwen3.8 27B | $0.425 → $0.420 (-$0.005) | $2.55 → $3.00 (+$0.45) | +$0.445 |
| NVIDIA: Nemotron 3 Ultra | $0.60 → $0.625 (+$0.025) | $2.40 → $3.125 (+$0.725) | +$0.75 |
| MoonshotAI Kimi Latest | $2.55 → $2.50 (-$0.05) | $12.75 → $14.00 (+$1.25) | +$1.20 |
| DeepSeek: DeepSeek V3.1 | $0.25 → $0.55 (+$0.30) | $0.95 → $1.65 (+$0.70) | +$1.00 |
| Qwen: Qwen2.5 VL 72B Instruct | $0.25 → $0.80 (+$0.55) | $0.75 → $1.00 (+$0.25) | +$0.80 |
| ReMM SLERP 13B | $0.45 → $0.35 (-$0.10) | $0.65 → $0.65 (0) | -$0.10 |
Added models
- inclusionai/ling-3.0-flash-fin – 262k ctx, prompt $0.06/1M, completion $0.18/1M.
- nvidia/nemotron-3.5-content-safety – 131k ctx, prompt & completion $0.20/1M each.
- x-ai/grok-4.3:batch – 1M ctx, prompt $1.00/1M, completion $2.00/1M (batch pricing).
No removals
Cheapest models today (per 1M tokens)
- IBM: Granite 4.0 Micro – prompt $0.017, completion $0.112
- Mistral: Mistral Nemo – prompt $0.019, completion $0.030
- Ling-3.0-flash – prompt $0.021, completion $0.063
Total models tracked: 427.
Originally published at The Token Ledger. Subscribe for the daily digest.
Top comments (0)