LLM price barometer

Every list-price move measured between 30 Aug – 6 Sept, biggest first. Percentages are on the blended 30/70 input/output price. Data snapshot: 2026-09-06.

Got more expensive (8)

• +69.8% DeepSeek V4 Pro 0813 (DeepSeek) — blended list price $1.58 to $2.69 per 1M tokens; input +69.8%, output +69.8%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• +54.9% DeepSeek V4 Pro 0423 (DeepSeek) — blended list price $0.74 to $1.15 per 1M tokens; input +54.9%, output +54.9%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• +51.9% Qwen3 235B A22B Instruct 2507 (Alibaba) — blended list price $0.27 to $0.41 per 1M tokens; input +2.9%, output +57.1%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• +49.0% Qwen3.5 397B A17B (Alibaba) — blended list price $1.75 to $2.61 per 1M tokens; input +41.0%, output +49.6%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• +40.5% Nemotron 3 Ultra (Nvidia) — blended list price $1.69 to $2.38 per 1M tokens; input +25.0%, output +42.0%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• +25.9% GLM 4.6 (Zhipu) — blended list price $1.35 to $1.71 per 1M tokens; input +27.9%, output +25.7%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• +16.4% Qwen3.8 27B (Alibaba) — blended list price $1.91 to $2.23 per 1M tokens; input -1.2%, output +17.6%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere. Input and output moved in opposite directions, so the effect on your bill depends on your token mix.
• +2.0% Qwen3.5-35B-A3B (Alibaba) — blended list price $0.95 to $0.97 per 1M tokens; input +25.0%, output 0.0%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.

Got cheaper (10)

• -64.2% Llama 3.3 70B Instruct (Meta) — blended list price $0.71 to $0.25 per 1M tokens; input -85.9%, output -54.9%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• -44.8% Qwen3.6 27B (Alibaba) — blended list price $2.70 to $1.49 per 1M tokens; input -50.0%, output -44.4%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• -41.6% DeepSeek V4 Flash 0731 (DeepSeek) — blended list price $0.15 to $0.08 per 1M tokens; input -23.1%, output -44.4%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• -25.0% Kimi K2.5 (Moonshot) — blended list price $2.28 to $1.71 per 1M tokens; input -25.0%, output -25.0%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• -18.8% GLM 5.2 (Zhipu) — blended list price $2.97 to $2.42 per 1M tokens; input -18.8%, output -18.8%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• -11.7% Llama 4 Maverick (Meta) — blended list price $0.62 to $0.55 per 1M tokens; input 0.0%, output -13.0%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• -11.4% Llama 4 Scout (Meta) — blended list price $0.27 to $0.24 per 1M tokens; input -9.1%, output -11.8%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.
• -9.8% DeepSeek V3 (DeepSeek) — blended list price $0.80 to $0.72 per 1M tokens; input +24.3%, output -13.5%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere. Input and output moved in opposite directions, so the effect on your bill depends on your token mix.
• -9.1% Devstral 2 2512 (Mistral) — blended list price $1.67 to $1.52 per 1M tokens; input -9.1%, output -9.1%.
• -7.5% Muse Glimmer 30B (meta) — blended list price $0.93 to $0.86 per 1M tokens; input 0.0%, output -8.3%. Not a repricing: this model is sold by several hosts at different rates, and the one we quote changed. The previous rate was still on sale elsewhere.

How these figures are calculated

• The window is real: it runs from the newest snapshot back to the oldest one within the preceding seven days, and those two dates are the ones named above. While the archive is young the window is shorter than a week.
• Prices are blended 30/70 input to output, the same figure the Efficiency score ranks on.
• List prices only. Batch and cache discounts are conditional on the customer implementing them.
• Corrections are excluded rather than annotated: when a published rate was wrong at the source and later fixed, the difference is a data correction, not a price move.
• A change of host is listed but never headlined: most models are sold by several hosts at different rates, this site quotes one of them, and when that host changes the figure moves although nobody repriced anything.
• A missing, zero or negative baseline yields no entry. New listings, renames and delistings get no percentage rather than an invented one.

The full cost-per-request formulas are in the cost per request methodology guide. For what to do about a price move, see the Cloud FinOps skill on GitHub.