AI infrastructure price compass

The price-performance compass for AI infrastructure — live LLM API costs, the daily Lodestar Index, and the cheapest GPU rentals. Updated automatically.

$0.084Lodestar Index / 1M

$0.015Cheapest blended

308Models tracked

7GPU types

Lodestar Index — USD / 1M tokens · production-grade (q≥80)

$0.084

cheapest model at quality ≥ 80: qwen/qwen3-30b-a3b-instruct-2507

history builds daily — the trend line appears after a few runs.

Model: qwen/qwen3-30b-a3b-instruct-2507

Cheapest blended LLM today: inclusionai/ling-2.6-flash at $0.015 / 1M tokens (blend 75% in / 25% out · 308 models)

Price movers (since last update)

No price changes since the last build. Movers light up the moment a provider repricing is detected.

LLM API price ranking (blended $/1M tokens)

Showing top 30 of 308 models by blended price.

#	Model	Provider	In $/1M	Out $/1M	Blended $/1M	Context
1	inclusionai/ling-2.6-flash	Inclusionai	0.01	0.03	0.015	262,144
2	meta-llama/llama-3.1-8b-instruct	Meta-Llama	0.02	0.03	0.022	131,072
3	mistralai/mistral-nemo	Mistralai	0.02	0.03	0.022	131,072
4	ibm-granite/granite-4.0-h-micro	Ibm-Granite	0.02	0.11	0.041	131,000
5	sao10k/l3-lunaris-8b	Sao10K	0.04	0.05	0.043	8,192
6	liquid/lfm-2-24b-a2b	Liquid	0.03	0.12	0.052	128,000
7	qwen/qwen-2.5-7b-instruct	Qwen	0.04	0.10	0.055	131,072
8	openai/gpt-oss-20b	Openai	0.03	0.14	0.057	131,072
9	mistralai/mistral-small-24b-instruct-2501	Mistralai	0.05	0.08	0.058	32,768
10	openai/gpt-oss-120b	Openai	0.03	0.15	0.060	131,072
11	gryphe/mythomax-l2-13b	Gryphe	0.06	0.06	0.060	4,096
12	amazon/nova-micro-v1	Amazon	0.04	0.14	0.061	128,000
13	ibm-granite/granite-4.1-8b	Ibm-Granite	0.05	0.10	0.062	131,072
14	google/gemma-3-4b-it	Google	0.05	0.10	0.062	131,072
15	cohere/command-r7b-12-2024	Cohere	0.04	0.15	0.066	128,000
16	meta-llama/llama-3.2-1b-instruct	Meta-Llama	0.03	0.20	0.070	131,072
17	arcee-ai/trinity-mini	Arcee-Ai	0.04	0.15	0.071	131,072
18	google/gemma-3n-e4b-it	Google	0.06	0.12	0.075	32,768
19	google/gemma-3-12b-it	Google	0.05	0.15	0.075	131,072
20	qwen/qwen3-30b-a3b-instruct-2507	Qwen	0.05	0.19	0.084	131,072
21	nvidia/nemotron-3-nano-30b-a3b	Nvidia	0.05	0.20	0.087	262,144
22	microsoft/phi-4	Microsoft	0.07	0.14	0.087	16,384
23	qwen/qwen3-235b-a22b-2507	Qwen	0.09	0.10	0.092	262,144
24	tencent/hy3-preview	Tencent	0.06	0.21	0.100	262,144
25	rekaai/reka-edge	Rekaai	0.10	0.10	0.100	16,384
26	mistralai/ministral-3b-2512	Mistralai	0.10	0.10	0.100	131,072
27	qwen/qwen3-235b-a22b-thinking-2507	Qwen	0.10	0.10	0.100	262,144
28	google/gemma-3-27b-it	Google	0.08	0.16	0.100	131,072
29	amazon/nova-lite-v1	Amazon	0.06	0.24	0.105	300,000
30	mistralai/mistral-small-3.2-24b-instruct	Mistralai	0.07	0.20	0.106	128,000

Price-performance frontier

Best value at each quality tier:
meta-llama/llama-3.1-8b-instruct ($0.022, q74) · qwen/qwen-2.5-7b-instruct ($0.055, q78) · qwen/qwen3-30b-a3b-instruct-2507 ($0.084, q84) · deepseek/deepseek-v3.2 ($0.257, q86) · openai/gpt-5.4-nano ($0.463, q95)

What a real job costs

Summarize 1,000 documents
cheapest: inclusionai/ling-2.6-flash $0.05 · priciest: openai/o1-pro $840.00 (16800.0x spread)

1M-token RAG pipeline
cheapest: inclusionai/ling-2.6-flash $0.01 · priciest: openai/o1-pro $240.00 (24000.0x spread)

10,000 chat turns
cheapest: inclusionai/ling-2.6-flash $0.22 · priciest: openai/o1-pro $3,900.00 (17727.3x spread)