The State of LLM Pricing
A live snapshot of large language model pricing across 17 providers and 185 models, rebuilt every day. Blended price weights input and output tokens 3:1 — a realistic mix for most applications. Free to cite and embed.
Updated 18 September 2026.
The headline
The gap between the cheapest and most expensive models is now roughly 3,750×. At the cheap end, Llama 3.2 1B Instruct costs $0.0100 per million blended tokens; at the top, GPT-4 costs $37.50. For most production workloads the interesting question isn't “what's cheapest” but “what's the cheapest model that's smart enough” — which is exactly what our value leaderboard answers.
Price momentum
See every price change →Since we began tracking, we’ve recorded 163 price changes — 102 drops vs 61 increases across 10 providers, plus 325 new models. The single biggest drop: Qwen3.5 Flash cached input down 93% on 10 August 2026.
The 10 cheapest models
| Llama 3.2 1B Instruct | Meta | $0.0100 | 1,055 |
| Llama 3.2 3b Instruct | Meta | $0.0200 | 1,109 |
| Llama 3.1 8b Instruct | Meta | $0.0250 | 1,186 |
| GPT OSS 20B | OpenAI | $0.0400 | 1,287 |
| gemma-3-4b-it | $0.0500 | 1,291 | |
| Gemma 3 4B IT (Bedrock) | AWS Bedrock | $0.0500 | 1,291 |
| GPT OSS 120B | OpenAI | $0.0590 | 1,366 |
| Amazon Nova Micro | AWS Bedrock | $0.0613 | 1,268 |
| Gemma 3 12B IT | $0.0625 | 1,334 | |
| Command R7B | Cohere | $0.0656 | 1,262 |
The 5 most expensive models
| GPT-4 | OpenAI | $37.50 | 1,186 |
| Claude Opus 4.1 (latest) | Anthropic | $30.00 | 1,418 |
| o1 | OpenAI | $26.25 | 1,353 |
| Claude 4 Opus | Anthropic | $25.75 | 1,366 |
| Claude Fable 5 (Bedrock) | AWS Bedrock | $20.00 | 1,493 |
Best value: intelligence per dollar
| Llama 3.2 1B Instruct | Meta | $0.0100 | 1,055 | 105,500 |
| Llama 3.2 3b Instruct | Meta | $0.0200 | 1,109 | 55,450 |
| Llama 3.1 8b Instruct | Meta | $0.0250 | 1,186 | 47,440 |
| GPT OSS 20B | OpenAI | $0.0400 | 1,287 | 32,175 |
| gemma-3-4b-it | $0.0500 | 1,291 | 25,820 | |
| Gemma 3 4B IT (Bedrock) | AWS Bedrock | $0.0500 | 1,291 | 25,820 |
| GPT OSS 120B | OpenAI | $0.0590 | 1,366 | 23,153 |
| Gemma 3 12B IT | $0.0625 | 1,334 | 21,344 | |
| Amazon Nova Micro | AWS Bedrock | $0.0613 | 1,268 | 20,702 |
| Command R7B | Cohere | $0.0656 | 1,262 | 19,230 |
Cite or embed this report
Download the JSON snapshot, grab the full intelligence dataset, or add a live “cheapest LLM today” badge to your own site. Data licensed under CC BY 4.0.
Prices from AI Calculator Pro, blended 3:1 input:output. Arena scores from LMArena / arena.ai. Rebuilt daily — last updated 18 September 2026.