Skip to main content
Radar

Every change to the models you run, with its source and its date. Releases, price changes, retirements, API changes, and incidents.Every change to the models you run, with its source.

Follow model changes
Citable dataset

LLM Pricing Statistics (2026)

Updated September 18, 2026 · Auto-generated from BenchLM's live dataset on every data refresh

Frontier LLM token prices are 84% below their March 2023 level as of September 2026, per the BenchLM Token Price Index (16 on a base of 100).

Frontier price drop since March 2023

Frontier LLM token prices are 84% below their March 2023 level as of September 2026, per the BenchLM Token Price Index (16 on a base of 100).

−84% (index: 16)

Price direction, last 12 months (frontier vs mid-tier)

Over the 12 months to September 2026, the BenchLM Token Price Index shows frontier LLM token prices up 72% year-over-year, while mid-tier prices went down 7.7%.

frontier +72%, mid-tier -7.7%

Models with tracked API pricing

As of September 18, 2026, BenchLM tracks live API pricing for 166 AI models.

166

Median price per 1M tokens (input / output)

As of September 18, 2026, the median LLM API price across 166 models tracked by BenchLM is $0.95 per 1M input tokens and $3.88 per 1M output tokens.

$0.95 / $3.88

Spread between cheapest and most expensive model

As of September 18, 2026, the most expensive LLM API tracked by BenchLM (o1-pro) costs roughly 4773x more per blended 1M tokens than the cheapest (Qwen3.7 Flash).

4773x

Cheapest frontier-tier model (top 10 overall)

As of September 18, 2026, the cheapest model in BenchLM's overall top 10 is Gemini 3.8 Flash at $0.75 per 1M input tokens and $3.75 per 1M output tokens.

Gemini 3.8 Flash ($0.75 in / $3.75 out)

Open-weight median discount vs proprietary

As of September 18, 2026, open-weight models on BenchLM have a median blended API price 74% lower than proprietary models ($0.53 vs $2.00 per 1M tokens at a 3:1 input:output ratio).

74% cheaper

Methodology & sources

Prices are USD per 1M tokens from BenchLM's pricing dataset (166 models, updated September 18, 2026). "Blended" price assumes a 3:1 input:output token ratio. "Frontier" means the top 10 models on BenchLM's overall ranking. Time-series figures come from the BenchLM Token Price Index (median blended price of flagship models, base March 2023 = 100).

Cite these statistics

Every number on this page is generated from BenchLM's live dataset and refreshed with each data update. Link any statistic directly using its anchor, or cite the page as:

BenchLM.ai, "LLM Statistics" (September 18, 2026), https://benchlm.ai/stats/llm-pricing

Questions

How much does an LLM API cost per million tokens?

As of September 18, 2026, the median price across 166 models tracked by BenchLM is $0.95 per 1M input tokens and $3.88 per 1M output tokens, with roughly a 4773x spread between the cheapest and most expensive models.

Are open-weight models cheaper than proprietary models?

Yes. As of September 18, 2026, open-weight models tracked by BenchLM have a median blended API price 74% lower than proprietary models ($0.53 vs $2.00 per 1M tokens).

What is the cheapest frontier-quality model?

As of September 18, 2026, the cheapest model in BenchLM's overall top 10 is Gemini 3.8 Flash, at $0.75 per 1M input tokens and $3.75 per 1M output tokens.

How much have LLM token prices dropped since GPT-4?

Frontier LLM token prices are 84% below their March 2023 (GPT-4 launch) level as of September 2026, per the BenchLM Token Price Index — though the last 12 months moved up, not down, at the frontier.

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.