The Token Ledger Digest – 2026-06-07

Cost‑impacting change

• Meta: Llama 3 8B Instruct – price increase: prompt from $0.04/1M to $0.14/1M; completion from $0.04/1M to $0.14/1M. Who should care: developers using this model for inference; cost per million tokens rises 250%.

Price changes

• Qwen: Qwen3.6 27B – prompt down slightly: $0.29/1M → $0.289/1M; completion down: $3.20/1M → $2.40/1M. Who should care: users of Qwen3.6 27B see modest savings, especially on completion‑heavy workloads.

Model removals

• Baidu: ERNIE 4.5 VL 28B A3B – removed; was $0.14/1M prompt, $0.56/1M completion. Who should care: anyone relying on this model must migrate.
• Arcee AI: Spotlight – removed; was $0.18/1M prompt, $0.18/1M completion. Who should care: users need to switch to an alternative.
• OpenAI: GPT-4 Turbo (older v1106) – removed; was $10.00/1M prompt, $30.00/1M completion. Who should care: teams still on the legacy preview must upgrade to a newer GPT‑4 variant.

Total models: 341.

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