DEV Community

4663437Mehdi
4663437Mehdi

Posted on Originally published at 4663437mehdi.github.io

The Token Ledger Digest – 2026-08-29

#ai #llm #api #news

The Token Ledger Digest – 2026-08-29

Lead: Qwen: Qwen3.8 2.4T A95B (batch) had the biggest cost impact, with prompt price rising from $2.00 → $2.50 per 1M tokens (+$0.50) and completion from $6.00 → $6.25 per 1M (+$0.25). Teams running batch workloads on this model should expect higher per‑token spend.

Added Models (10)

  • Z.ai: GLM 5.3 Flash (batch) – prompt $0.15, completion $0.50 /1M.
  • DeepSeek: V4 Pro 0813 (batch) – prompt $1.32, completion $3.96 /1M.
  • Meta: Muse Glimmer 30B (batch) – prompt $0.35, completion $1.50 /1M.
  • DeepSeek: V4 Flash 0731 (batch) – prompt $0.14, completion $0.28 /1M.
  • Thinking Machines: Inkling Small (batch) – prompt $0.50, completion $1.20 /1M.
  • MoonshotAI: Kimi K3 (batch) – prompt $3.00, completion $15.00 /1M.
  • Google: Gemma 4 31B (batch) – prompt $0.39, completion $0.97 /1M.
  • Qwen: Qwen3.5‑9B (batch) – prompt $0.17, completion $0.25 /1M.
  • OpenAI: gpt‑oss‑120b (batch) – prompt $0.15, completion $0.60 /1M.
  • OpenAI: gpt‑oss‑20b (batch) – prompt $0.05, completion $0.20 /1M. Who should care: developers evaluating new batch‑enabled options for cost‑sensitive inference.

Removed Models (2)

  • AllenAI: Olmo 3 32B Think – previously prompt $0.15, completion $0.50 /1M.
  • Arcee AI: Virtuoso Large – previously prompt $0.75, completion $1.20 /1M. Who should care: users of these models must migrate to alternatives.

Price Changes (7)

  • Z.ai: GLM Latest – prompt $1.25 → $1.19 (−$0.06), completion $4.40 → $4.18 (−$0.22) /1M.
  • Qwen: Qwen3.8 2.4T A95B (batch) – see lead.
  • Meta: Muse Glimmer 30B – prompt $0.35 → $0.30 (−$0.05), completion $1.50 → $1.20 (−$0.30) /1M.
  • DeepSeek: V4 Flash 0731 – prompt $0.06 → $0.045 (−$0.015), completion $0.12 → $0.09 (−$0.03) /1M.
  • Tencent: Hy3 – prompt $0.0825 → $0.132 (+$0.05), completion $0.33 → $0.528 (+$0.20) /1M.
  • DeepSeek: V4 Pro 0423 – prompt $0.7416 → $0.6099 (−$0.13), completion $1.4832 → $1.2197 (−$0.26) /1M.
  • DeepSeek: V4 Flash 0423 – prompt $0.0868 → $0.0832 (−$0.004), completion $0.1736 → $0.1663 (−$0.01) /1M. Who should care: existing users of these models should adjust cost forecasts accordingly.

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

Top comments (0)