Token Ledger Digest – 2026-07-31
Added
- DeepSeek: DeepSeek V4 Flash 0731 – New model with 1M‑token context. Prompt $0.14/1M, completion $0.28/1M. Who should care: Teams needing long‑context generation at low cost.
Removed
- OpenAI: o3 Deep Research – Prompt $10.00/1M, completion $40.00/1M.
- OpenAI: o4 Mini Deep Research – Prompt $2.00/1M, completion $8.00/1M.
- OpenAI: GPT-5 Codex – Prompt $1.25/1M, completion $10.00/1M. Who should care: Any workflows relying on these models must migrate to alternatives.
Price Changes
| Model | Old Prompt ($/1M) | New Prompt ($/1M) | Δ Prompt | Old Compl. ($/1M) | New Compl. ($/1M) | Δ Compl. | Who should care |
|---|---|---|---|---|---|---|---|
| Poolside: Laguna S 2.1 | 0.10 | 0.09 | –0.01 | 0.20 | 0.18 | –0.02 | Cost‑sensitive inference on Laguna S. |
| OpenAI: GPT-5.6 Luna Pro | 0.50 | 0.10 | –0.40 | 3.00 | 0.60 | –2.40 | Heavy users of Luna Pro see ~80% cost cut. |
| OpenAI: GPT-5.6 Luna | 0.50 | 0.10 | –0.40 | 3.00 | 0.60 | –2.40 | Same impact as Luna Pro. |
| OpenAI: GPT-5.6 Terra Pro | 1.25 | 1.00 | –0.25 | 7.50 | 6.00 | –1.50 | Terra Pro workloads become cheaper. |
| OpenAI: GPT-5.6 Terra | 1.25 | 1.00 | –0.25 | 7.50 | 6.00 | –1.50 | Same as Terra Pro. |
| Z.ai: GLM 5.2 | 0.70 | 1.23 | +0.53 | 2.20 | 3.87 | +1.67 | GLM 5.2 now significantly more expensive. |
| NVIDIA: Nemotron 3 Ultra | 0.50 | 0.60 | +0.10 | 2.20 | 3.60 | +1.40 | Ultra‑scale users face higher spend. |
| MoonshotAI Kimi Latest | 2.90 | 2.90 | 0.00 | 15.00 | 14.00 | –1.00 | Slight completion‑cost relief. |
| MoonshotAI: Kimi K2.6 | 0.65 | 0.95 | +0.30 | 2.72 | 4.00 | +1.28 | Kimi K2.6 cost rises noticeably. |
| Qwen: Qwen3 VL 30B A3B Instruct | 0.15 | 0.13 | –0.02 | 0.60 | 0.52 | –0.08 | Minor savings for vision‑language tasks. |
| Qwen: Qwen3 235B A22B Thinking 2507 | 0.30 | 0.23 | –0.07 | 3.00 | 2.30 | –0.70 | Thinking model becomes cheaper. |
Most cost‑impacting change: The OpenAI GPT-5.6 Luna Pro and Luna models dropped prompt pricing from $0.50 to $0.10/1M and completion from $3.00 to $0.60/1M – an ~80% reduction, saving up to $2.80 per million tokens.
Cheapest Models Today
- inclusionAI: Ling-2.6-flash – $0.01/1M prompt, $0.03/1M completion
- IBM: Granite 4.0 Micro – $0.017/1M prompt, $0.112/1M completion
- Mistral: Mistral Nemo – $0.019/1M prompt, $0.03/1M completion
Total models tracked: 365.
Originally published at The Token Ledger. Subscribe for the daily digest.
Top comments (0)