DEV Community

4663437Mehdi
4663437Mehdi

Posted on Originally published at 4663437mehdi.github.io

The Token Ledger Digest – 2026-08-23

#ai #llm #api #news

The Token Ledger Digest – 2026-08-23

Biggest cost impact: MoonshotAI’s Kimi K2.6 saw its completion price jump from $2.28 / 1M to $4.00 / 1M (+$1.72 / 1M) and its prompt price rise from $0.54 / 1M to $0.95 / 1M (+$0.41 / 1M). Developers using this model for long‑form generation should reassess budget forecasts.

Price changes

Model Change Old → New (per 1M) Who should care
DeepSeek V4 Flash Vision Exp Prompt ↓, Completion ↓ Prompt: $0.44 → $0.22; Completion: $1.32 → $0.66 Vision‑heavy workloads see ~50% cost cut.
Qwen3.8 27B Prompt ↓, Completion ↓ Prompt: $0.45 → $0.40; Completion: $3.20 → $3.00 Moderate savings for mid‑size models.
DeepSeek V4 Pro 0813 Prompt ↓, Completion ↓ Prompt: $1.188 → $1.122; Completion: $3.564 → $3.366 Slight trim for Pro tier users.
Meta Muse Glimmer 30B Prompt ↑, Completion ↑ Prompt: $0.30 → $0.35; Completion: $1.10 → $1.50 Cost rise; consider alternatives for high‑volume tasks.
DeepSeek V4 Flash Latest Prompt ↓, Completion ↓ Prompt: $0.065 → $0.040; Completion: $0.18 → $0.13 Ultra‑cheap flash model gets even cheaper.
Thinking Machines Inkling Prompt ↑, Completion ↔ Prompt: $0.95 → $1.00; Completion: $4.05 unchanged Prompt‑heavy apps face modest increase.
DeepSeek V4 Pro 0423 Prompt ↓, Completion ↓ Prompt: $0.4475 → $0.3969; Completion: $0.8951 → $0.7938 Continued cost reduction for Pro line.
DeepSeek V4 Flash 0423 Prompt ↓, Completion ↓ Prompt: $0.0713 → $0.0517; Completion: $0.1425 → $0.1033 Flash tier gets cheaper across the board.
OpenAI gpt‑oss‑120b Prompt ↑, Completion ↔ Prompt: $0.030 → $0.037; Completion: $0.17 unchanged Small prompt‑cost uptick; negligible for most.
Qwen3 30B A3B Prompt ↓, Completion ↓ Prompt: $0.13 → $0.12; Completion: $0.52 → $0.50 Minor savings for this variant.

Added model

  • Tencent: Hy‑MT2‑7B – Prompt $0.074 / 1M, Completion $0.295 / 1M, context 8192. Suitable for low‑latency, budget‑conscious tasks.

Cheapest models today (per 1M tokens)

  1. inclusionAI: Ling‑2.6‑flash – Prompt $0.01, Completion $0.03
  2. IBM: Granite 4.0 Micro – Prompt $0.017, Completion $0.112
  3. Mistral: Mistral Nemo – Prompt $0.019, Completion $0.03

Total models tracked: 422. No other meaningful changes reported.

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

Top comments (0)