The Token Ledger – 2026-08-13
Lead change: DeepSeek’s V4 Pro saw its pricing roughly double. Prompt price rose from $0.63 to $1.17 per 1M tokens (+85%); completion price rose from $1.26 to $2.34 per 1M tokens (+86%). Teams budgeting for DeepSeek V4 Pro should revisit cost forecasts.
Added models
| Model | Prompt ($/1M) | Completion ($/1M) | Who should care |
|---|---|---|---|
| ByteDance Seed: Seed 2.1 Turbo | 0.50 | 2.50 | Users needing large‑context (262k) with moderate cost. |
| Qwen: Qwen3.8 2.4T A95B | 2.00 | 6.00 | Applications that can afford higher cost for 1M‑token context. |
| DeepSeek: DeepSeek V4 Pro 0813 | 0.435 | 0.87 | Cost‑conscious users wanting a 1M‑token window. |
| SpaceXAI: Grok 4.6 | 2.00 | 6.00 | Similar to Qwen3.8; suited for long‑form tasks. |
Removed model
- inclusionAI: Ling 3.0 Tiny (free) – prompt $0, completion $0. Users relying on the free tier must switch to a paid alternative.
Price changes
| Model | Prompt change ($/1M) | Completion change ($/1M) | Who should care |
|---|---|---|---|
| MoonshotAI: Kimi K2.7 Code | ↓ 0.70 → 0.67 | ↓ 3.50 → 3.40 | Slight savings for code‑gen workloads. |
| DeepSeek: DeepSeek V4 Pro | ↑ 0.63 → 1.17 | ↑ 1.26 → 2.34 | Significant cost increase; see lead. |
| Qwen: Qwen3.5‑35B‑A3B | ↑ 0.14 → 0.25 | ↑ 1.00 → 1.25 | Moderate rise for mid‑size models. |
| Z.ai: GLM 4.6 | ↓ 0.55 → 0.50 | ↓ 2.20 → 2.00 | Small reduction for general‑purpose use. |
Total models tracked: 409.
Originally published at The Token Ledger. Subscribe for the daily digest.
Top comments (0)