Is it cheaper to self-host or use an API?

At low volume, API is cheaper. At very high volume (24/7 high utilization), self-hosting wins. The crossover point is usually around 10 to 100 million tokens per day depending on model size. Our self-host estimates above assume on-demand cloud GPUs.

Where do I download the weights?

HuggingFace is the canonical source for nearly every OSS model. Check the model detail page on BenchGecko for the exact HF ID.

Which OSS models are safe for commercial use?

Apache 2.0 and MIT licensed models (Mistral, most Qwen versions, Gemma) are clean for commercial use. Llama Community License has revenue-threshold restrictions. Always read the specific license before shipping.

Can I fine-tune an OSS model?

Yes, if the license permits (most do). Fine-tuning services on Together, Fireworks, and DeepInfra run $1 to $5 per million training tokens. Deploy the fine-tune back on the same platform or download the adapter.

Licensing · Open source

Open source LLM pricing

Every open-source LLM with license, cheapest API provider, and rough self-host cost. Weights on HuggingFace, API on OpenRouter, DeepInfra, Together, Fireworks, or self-host.

Models243

Cheapest API$0.00

LicensesApache · MIT · Custom

All pricing Pricing home

What this page is

This page lists every open-source LLM with priced API access. For each, we show the license, cheapest API on the market, and a rough self-host cost estimate based on parameter count. Weights are downloadable from HuggingFace. APIs are offered by OpenRouter, DeepInfra, Together, Fireworks, Groq, Cerebras, and others. For very high volume, self-hosting on reserved GPUs is cheaper; for anything under 100M tokens per day, API access usually wins on total cost.

Open source models · ranked by API price

Cheapest input price first.

#	Model	License	Cheapest API	Params	Self-host est	Weights
1	Gemma 3 12B (free)	Gemma License	$0.00/M	-	-	HF
2	Gemma 3 27B (free)	Gemma License	$0.00/M	27B	~$1.50/hr (1x A100)	HF
3	Gemma 3 4B (free)	Gemma License	$0.00/M	-	-	HF
4	Gemma 3n 2B (free)	Gemma License	$0.00/M	-	-	HF
5	Gemma 3n 4B (free)	Gemma License	$0.00/M	-	-	HF
6	Gemma 4 26B A4B (free)	Apache 2.0	$0.00/M	27B	~$1.50/hr (1x A100)	HF
7	Gemma 4 31B (free)	Apache 2.0	$0.00/M	33B	~$1.50/hr (1x A100)	HF
8	GLM 4.5 Air (free)	MIT	$0.00/M	110B	~$12/hr (4x A100)	HF
9	gpt-oss-120b (free)	Apache 2.0	$0.00/M	120B	~$12/hr (4x A100)	HF
10	gpt-oss-20b (free)	Apache 2.0	$0.00/M	22B	~$1.50/hr (1x A100)	HF
11	Hermes 3 405B Instruct (free)	Llama Community	$0.00/M	-	-	HF
12	Laguna M.1 (free)	Custom OSS	$0.00/M	-	-	-
13	Laguna XS.2 (free)	Custom OSS	$0.00/M	-	-	-
14	LFM2.5-1.2B-Instruct (free)	other	$0.00/M	1B	~$0.50/hr (1x L4)	HF
15	LFM2.5-1.2B-Thinking (free)	other	$0.00/M	1B	~$0.50/hr (1x L4)	HF
16	Llama 3.2 3B Instruct (free)	Llama Community	$0.00/M	3B	~$0.50/hr (1x L4)	HF
17	Llama 3.3 70B Instruct (free)	Llama Community	$0.00/M	71B	~$4/hr (2x A100)	HF
18	MiniMax M2.5 (free)	other	$0.00/M	229B	~$30/hr (8x H100)	HF
19	Mistral Small 3.1 24B (free)	Custom OSS	$0.00/M	-	-	-
20	Nemotron 3 Nano 30B A3B (free)	other	$0.00/M	32B	~$1.50/hr (1x A100)	HF
21	Nemotron 3 Nano Omni (free)	Custom OSS	$0.00/M	-	-	-
22	Nemotron 3 Super (free)	other	$0.00/M	124B	~$12/hr (4x A100)	HF
23	Nemotron 3 Ultra (free)	Custom OSS	$0.00/M	-	-	-
24	Nemotron 3.5 Content Safety (free)	Custom OSS	$0.00/M	-	-	-
25	Nemotron Nano 12B 2 VL (free)	other	$0.00/M	13B	~$1.50/hr (1x A100)	HF
26	Nemotron Nano 9B V2 (free)	other	$0.00/M	9B	~$0.50/hr (1x L4)	HF
27	Nex-N2-Pro (free)	Custom OSS	$0.00/M	-	-	-
28	North Mini Code (free)	Custom OSS	$0.00/M	-	-	-
29	Qwen3 4B (free)	Custom OSS	$0.00/M	-	-	-
30	Qwen3 Coder 480B A35B (free)	Apache 2.0	$0.00/M	480B	~$30/hr (8x H100)	HF
31	Qwen3 Next 80B A3B Instruct (free)	Apache 2.0	$0.00/M	81B	~$12/hr (4x A100)	HF
32	Qwen3.6 Plus Preview (free)	Custom OSS	$0.00/M	-	-	-
33	Step 3.5 Flash (free)	Custom OSS	$0.00/M	-	-	-
34	Trinity Large Preview (free)	Apache 2.0	$0.00/M	399B	~$30/hr (8x H100)	HF
35	Trinity Mini (free)	Apache 2.0	$0.00/M	-	-	HF
36	Uncensored (free)	Apache 2.0	$0.00/M	24B	~$1.50/hr (1x A100)	HF
37	LFM2-2.6B	Custom OSS	$0.01/M	-	-	-
38	LFM2-8B-A1B	Custom OSS	$0.01/M	-	-	-
39	Granite 4.0 Micro	Apache 2.0	$0.02/M	3B	~$0.50/hr (1x L4)	HF
40	Llama 3.1 8B Instruct	Llama Community	$0.02/M	8B	~$0.50/hr (1x L4)	HF
41	Mistral Nemo	Apache 2.0	$0.02/M	-	-	HF
42	Llama 3.2 1B Instruct	Llama Community	$0.03/M	1B	~$0.50/hr (1x L4)	HF
43	gpt-oss-20b	Apache 2.0	$0.03/M	22B	~$1.50/hr (1x A100)	HF
44	Gemma 2 9B	Gemma License	$0.03/M	9B	~$0.50/hr (1x L4)	HF
45	LFM2-24B-A2B	other	$0.03/M	24B	~$1.50/hr (1x A100)	HF
46	Qwen2.5 Coder 7B Instruct	Apache 2.0	$0.03/M	8B	~$0.50/hr (1x L4)	HF
47	Qwen-Turbo	Custom OSS	$0.03/M	-	-	-
48	gpt-oss-120b	Apache 2.0	$0.04/M	120B	~$12/hr (4x A100)	HF
49	Llama 3 8B Lunaris	Llama Community	$0.04/M	8B	~$0.50/hr (1x L4)	HF
50	Nemotron Nano 9B V2	other	$0.04/M	9B	~$0.50/hr (1x L4)	HF
51	Qwen2.5 7B Instruct	Apache 2.0	$0.04/M	8B	~$0.50/hr (1x L4)	HF
52	Trinity Mini	Apache 2.0	$0.04/M	-	-	HF
53	Qwen3 30B A3B Instruct 2507	Apache 2.0	$0.05/M	-	-	HF
54	Gemma 3 12B	Gemma License	$0.05/M	-	-	HF
55	Gemma 3 4B	Gemma License	$0.05/M	-	-	HF
56	Granite 4.1 8B	Custom OSS	$0.05/M	-	-	-
57	Llama 3.2 3B Instruct	Llama Community	$0.05/M	3B	~$0.50/hr (1x L4)	HF
58	Mistral Small 3	Apache 2.0	$0.05/M	-	-	HF
59	Nemotron 3 Nano 30B A3B	other	$0.05/M	32B	~$1.50/hr (1x A100)	HF
60	Olmo 2 32B Instruct	Apache 2.0	$0.05/M	32B	~$1.50/hr (1x A100)	HF
61	Gemma 3n 4B	Gemma License	$0.06/M	-	-	HF
62	Gemma 4 26B A4B	Apache 2.0	$0.06/M	27B	~$1.50/hr (1x A100)	HF
63	GLM 4.7 Flash	MIT	$0.06/M	31B	~$1.50/hr (1x A100)	HF
64	MythoMax 13B	other	$0.06/M	-	-	HF
65	Hy3 preview	Custom OSS	$0.06/M	-	-	-
66	Qwen3.5-Flash	Custom OSS	$0.07/M	-	-	-
67	ERNIE 4.5 21B A3B	Apache 2.0	$0.07/M	-	-	HF
68	ERNIE 4.5 21B A3B Thinking	Apache 2.0	$0.07/M	22B	~$1.50/hr (1x A100)	HF
69	Phi 4	MIT	$0.07/M	-	-	HF
70	Qwen3 Coder 30B A3B Instruct	Apache 2.0	$0.07/M	31B	~$1.50/hr (1x A100)	HF
71	gpt-oss-safeguard-20b	Apache 2.0	$0.07/M	-	-	HF
72	Mistral Small 3.2 24B	Apache 2.0	$0.07/M	-	-	HF
73	Gemma 3 27B	Gemma License	$0.08/M	27B	~$1.50/hr (1x A100)	HF
74	Nemotron 3 Super	other	$0.08/M	124B	~$12/hr (4x A100)	HF
75	Phi 4 Mini Instruct	Custom OSS	$0.08/M	-	-	-
76	Qwen3 32B	Apache 2.0	$0.08/M	-	-	HF
77	DeepSeek V4 Flash	Custom OSS	$0.08/M	-	-	-
78	MiMo-V2-Flash	MIT	$0.09/M	310B	~$30/hr (8x H100)	HF
79	Qwen3 235B A22B Instruct 2507	Apache 2.0	$0.09/M	-	-	HF
80	Qwen3 Next 80B A3B Instruct	Apache 2.0	$0.09/M	81B	~$12/hr (4x A100)	HF
81	Tongyi DeepResearch 30B A3B	Apache 2.0	$0.09/M	31B	~$1.50/hr (1x A100)	HF
82	Qwen3 Next 80B A3B Thinking	Apache 2.0	$0.10/M	-	-	HF
83	Devstral Small 1.1	Apache 2.0	$0.10/M	-	-	HF
84	Laguna XS.2	Custom OSS	$0.10/M	-	-	-
85	Llama 3.3 70B Instruct	Llama Community	$0.10/M	71B	~$4/hr (2x A100)	HF
86	Llama 4 Scout	other	$0.10/M	-	-	HF
87	Ministral 3 3B 2512	Apache 2.0	$0.10/M	-	-	HF
88	Mistral Small Creative	Custom OSS	$0.10/M	-	-	-
89	Qwen3 14B	Apache 2.0	$0.10/M	-	-	HF
90	Qwen3.5-9B	Apache 2.0	$0.10/M	10B	~$0.50/hr (1x L4)	HF
91	Reka Edge	other	$0.10/M	7B	~$0.50/hr (1x L4)	HF
92	Reka Flash 3	Apache 2.0	$0.10/M	21B	~$1.50/hr (1x A100)	HF
93	Step 3.5 Flash	Apache 2.0	$0.10/M	199B	~$12/hr (4x A100)	HF
94	UI-TARS 7B	Apache 2.0	$0.10/M	8B	~$0.50/hr (1x L4)	HF
95	Voxtral Small 24B 2507	Apache 2.0	$0.10/M	24B	~$1.50/hr (1x A100)	HF
96	Qwen3 VL 32B Instruct	Apache 2.0	$0.10/M	33B	~$1.50/hr (1x A100)	HF
97	MiMo-V2.5	Custom OSS	$0.10/M	-	-	-
98	Mistral 7B Instruct v0.1	Apache 2.0	$0.11/M	-	-	HF
99	Qwen3 Coder Next	Apache 2.0	$0.11/M	80B	~$4/hr (2x A100)	HF
100	Qwen3 8B	Apache 2.0	$0.12/M	8B	~$0.50/hr (1x L4)	HF
101	Qwen3 VL 8B Instruct	Apache 2.0	$0.12/M	9B	~$0.50/hr (1x L4)	HF
102	Qwen3 VL 8B Thinking	Apache 2.0	$0.12/M	9B	~$0.50/hr (1x L4)	HF
103	Gemma 4 31B	Apache 2.0	$0.12/M	33B	~$1.50/hr (1x A100)	HF
104	Qwen3 30B A3B	Apache 2.0	$0.12/M	-	-	HF
105	GLM 4.5 Air	MIT	$0.13/M	110B	~$12/hr (4x A100)	HF
106	Hermes 4 70B	Llama Community	$0.13/M	-	-	HF
107	Qwen3 30B A3B Thinking 2507	Apache 2.0	$0.13/M	31B	~$1.50/hr (1x A100)	HF
108	Qwen3 VL 30B A3B Instruct	Apache 2.0	$0.13/M	-	-	HF
109	Qwen3 VL 30B A3B Thinking	Apache 2.0	$0.13/M	31B	~$1.50/hr (1x A100)	HF
110	DeepSeek V3.1 Nex N1	Apache 2.0	$0.14/M	671B	~$60+/hr (multi-node)	HF
111	Qwen VL Plus	Custom OSS	$0.14/M	-	-	-
112	ERNIE 4.5 VL 28B A3B	Apache 2.0	$0.14/M	29B	~$1.50/hr (1x A100)	HF
113	Hermes 2 Pro - Llama-3 8B	Llama Community	$0.14/M	8B	~$0.50/hr (1x L4)	HF
114	Hunyuan A13B Instruct	other	$0.14/M	-	-	HF
115	Llama 3 8B Instruct	Llama Community	$0.14/M	8B	~$0.50/hr (1x L4)	HF
116	Qwen3.5-35B-A3B	Apache 2.0	$0.14/M	36B	~$4/hr (2x A100)	HF
117	Qwen3.6 35B A3B	Custom OSS	$0.14/M	-	-	-
118	Qwen3 235B A22B Thinking 2507	Apache 2.0	$0.15/M	-	-	HF
119	Llama 4 Maverick	other	$0.15/M	402B	~$30/hr (8x H100)	HF
120	MiniMax M2.5	other	$0.15/M	229B	~$30/hr (8x H100)	HF
121	Ministral 3 8B 2512	Apache 2.0	$0.15/M	9B	~$0.50/hr (1x L4)	HF
122	Mistral Small 4	Apache 2.0	$0.15/M	119B	~$12/hr (4x A100)	HF
123	Olmo 3 32B Think	Apache 2.0	$0.15/M	1M	~$0.50/hr (1x L4)	HF
124	Olmo 3.1 32B Think	Custom OSS	$0.15/M	-	-	-
125	QwQ 32B	Apache 2.0	$0.15/M	33B	~$1.50/hr (1x A100)	HF
126	Rnj 1 Instruct	Apache 2.0	$0.15/M	8B	~$0.50/hr (1x L4)	HF
127	Llama Guard 4 12B	other	$0.18/M	-	-	HF
128	Qwen3.6 Flash	Custom OSS	$0.19/M	-	-	-
129	Qwen3 Coder Flash	Custom OSS	$0.20/M	-	-	-
130	Qwen3.5-27B	Apache 2.0	$0.20/M	28B	~$1.50/hr (1x A100)	HF
131	INTELLECT-3	MIT	$0.20/M	107B	~$12/hr (4x A100)	HF
132	Laguna M.1	Custom OSS	$0.20/M	-	-	-
133	LongCat Flash Chat	MIT	$0.20/M	562B	~$60+/hr (multi-node)	HF
134	MiniMax-01	Custom OSS	$0.20/M	-	-	HF
135	Ministral 3 14B 2512	Apache 2.0	$0.20/M	-	-	HF
136	Nemotron Nano 12B 2 VL	other	$0.20/M	13B	~$1.50/hr (1x A100)	HF
137	Olmo 3.1 32B Instruct	Apache 2.0	$0.20/M	-	-	HF
138	Qwen2.5 VL 32B Instruct	Apache 2.0	$0.20/M	33B	~$1.50/hr (1x A100)	HF
139	Qwen3 VL 235B A22B Instruct	Apache 2.0	$0.20/M	236B	~$30/hr (8x H100)	HF
140	Saba	Custom OSS	$0.20/M	-	-	-
141	Step 3.7 Flash	Custom OSS	$0.20/M	-	-	-
142	DeepSeek V3	Custom OSS	$0.20/M	685B	~$60+/hr (multi-node)	HF
143	DeepSeek V3.1	MIT	$0.21/M	-	-	HF
144	DeepSeek V3.2	MIT	$0.21/M	685B	~$60+/hr (multi-node)	HF
145	Qwen3 Coder 480B A35B	Apache 2.0	$0.22/M	480B	~$30/hr (8x H100)	HF
146	DeepSeek V3 0324	MIT	$0.24/M	685B	~$60+/hr (multi-node)	HF
147	MiniMax M2.7	Custom OSS	$0.24/M	-	-	-
148	Rocinante 12B	other	$0.25/M	-	-	HF
149	Trinity Large Thinking	Apache 2.0	$0.25/M	399B	~$30/hr (8x H100)	HF
150	MiniMax M2	other	$0.26/M	229B	~$30/hr (8x H100)	HF
151	Qwen Plus 0728	Custom OSS	$0.26/M	-	-	-
152	Qwen Plus 0728 (thinking)	Custom OSS	$0.26/M	-	-	-
153	Qwen-Plus	Custom OSS	$0.26/M	-	-	-
154	Qwen3 VL 235B A22B Thinking	Apache 2.0	$0.26/M	236B	~$30/hr (8x H100)	HF
155	Qwen3.5 Plus 2026-02-15	Custom OSS	$0.26/M	-	-	-
156	Qwen3.5-122B-A10B	Apache 2.0	$0.26/M	125B	~$12/hr (4x A100)	HF
157	DeepSeek V3.1 Terminus	MIT	$0.27/M	-	-	HF
158	DeepSeek V3.2 Exp	MIT	$0.27/M	-	-	HF
159	ERNIE 4.5 300B A47B	Apache 2.0	$0.28/M	-	-	HF
160	Qwen3.6 27B	Custom OSS	$0.28/M	-	-	-
161	R1 Distill Qwen 32B	MIT	$0.29/M	33B	~$1.50/hr (1x A100)	HF
162	Codestral 2508	Custom OSS	$0.30/M	-	-	-
163	Cydonia 24B V4.1	Custom OSS	$0.30/M	24B	~$1.50/hr (1x A100)	HF
164	DeepSeek R1T2 Chimera	MIT	$0.30/M	-	-	HF
165	GLM 4.6V	MIT	$0.30/M	-	-	HF
166	MiniMax M2.1	other	$0.30/M	229B	~$30/hr (8x H100)	HF
167	MiniMax M3	Custom OSS	$0.30/M	-	-	-
168	Qwen3.5 Plus 2026-04-20	Custom OSS	$0.30/M	-	-	-
169	Qwen3.7 Plus	Custom OSS	$0.32/M	-	-	-
170	Qwen3.6 Plus	Custom OSS	$0.33/M	-	-	-
171	Llama 3.2 11B Vision Instruct	Llama Community	$0.34/M	11B	~$1.50/hr (1x A100)	HF
172	Mistral Small 3.1 24B	Apache 2.0	$0.35/M	-	-	HF
173	Qwen2.5 72B Instruct	other	$0.36/M	73B	~$4/hr (2x A100)	HF
174	Kimi K2.5	other	$0.38/M	1.1T	~$60+/hr (multi-node)	HF
175	Qwen3.5 397B A17B	Apache 2.0	$0.39/M	403B	~$30/hr (8x H100)	HF
176	DeepSeek V3.2 Speciale	MIT	$0.40/M	-	-	HF
177	Devstral 2 2512	other	$0.40/M	125B	~$12/hr (4x A100)	HF
178	Devstral Medium	Custom OSS	$0.40/M	-	-	-
179	GLM 4.7	MIT	$0.40/M	358B	~$30/hr (8x H100)	HF
180	Llama 3.1 70B Instruct	Llama Community	$0.40/M	71B	~$4/hr (2x A100)	HF
181	Llama 3.3 Nemotron Super 49B V1.5	other	$0.40/M	50B	~$4/hr (2x A100)	HF
182	Mistral Medium 3	Custom OSS	$0.40/M	-	-	-
183	Mistral Medium 3.1	Custom OSS	$0.40/M	-	-	-
184	UnslopNemo 12B	Custom OSS	$0.40/M	12B	~$1.50/hr (1x A100)	HF
185	ERNIE 4.5 VL 424B A47B	Apache 2.0	$0.42/M	424B	~$30/hr (8x H100)	HF
186	GLM 4.6	MIT	$0.43/M	357B	~$30/hr (8x H100)	HF
187	DeepSeek V4 Pro	Custom OSS	$0.43/M	-	-	-
188	MiMo-V2.5-Pro	Custom OSS	$0.43/M	-	-	-
189	ReMM SLERP 13B	CC BY	$0.45/M	-	-	HF
190	Qwen3 235B A22B	Apache 2.0	$0.46/M	235B	~$30/hr (8x H100)	HF
191	Llama Guard 3 8B	Llama Community	$0.48/M	8B	~$0.50/hr (1x L4)	HF
192	Mistral Large 3 2512	Custom OSS	$0.50/M	-	-	-
193	Nemotron 3 Ultra	Custom OSS	$0.50/M	-	-	-
194	Nex-N2-Pro	Custom OSS	$0.50/M	-	-	-
195	R1 0528	MIT	$0.50/M	685B	~$60+/hr (multi-node)	HF
196	Llama 3 70B Instruct	Llama Community	$0.51/M	71B	~$4/hr (2x A100)	HF
197	Qwen VL Max	Custom OSS	$0.52/M	-	-	-
198	Mixtral 8x7B Instruct	Apache 2.0	$0.54/M	47B	~$4/hr (2x A100)	HF
199	Skyfall 36B V2	other	$0.55/M	37B	~$4/hr (2x A100)	HF
200	Kimi K2 0711	other	$0.57/M	1.0T	~$60+/hr (multi-node)	HF
201	GLM 4.5	MIT	$0.60/M	358B	~$30/hr (8x H100)	HF
202	GLM 4.5V	MIT	$0.60/M	108B	~$12/hr (4x A100)	HF
203	GLM 5	MIT	$0.60/M	754B	~$60+/hr (multi-node)	HF
204	Kimi K2 0905	other	$0.60/M	1.0T	~$60+/hr (multi-node)	HF
205	Kimi K2 Thinking	other	$0.60/M	1.1T	~$60+/hr (multi-node)	HF
206	Llama 3.1 Nemotron Ultra 253B v1	other	$0.60/M	-	-	HF
207	Kimi K2.7 Code	Custom OSS	$0.61/M	-	-	-
208	WizardLM-2 8x22B	Custom OSS	$0.62/M	-	-	-
209	Gemma 2 27B	Gemma License	$0.65/M	27B	~$1.50/hr (1x A100)	HF
210	Llama 3.3 Euryale 70B	Llama Community	$0.65/M	71B	~$4/hr (2x A100)	HF
211	Qwen3 Coder Plus	Custom OSS	$0.65/M	-	-	-
212	Kimi K2.6	Custom OSS	$0.66/M	-	-	-
213	Qwen2.5 Coder 32B Instruct	Apache 2.0	$0.66/M	33B	~$1.50/hr (1x A100)	HF
214	Aion-1.0-Mini	Apache 2.0	$0.70/M	-	-	HF
215	Hermes 3 70B Instruct	Llama Community	$0.70/M	71B	~$4/hr (2x A100)	HF
216	R1	MIT	$0.70/M	685B	~$60+/hr (multi-node)	HF
217	Qwen3 Max	Custom OSS	$0.78/M	-	-	-
218	Qwen3 Max Thinking	Custom OSS	$0.78/M	-	-	-
219	CodeLLaMa 7B Instruct Solidity	Llama Community	$0.80/M	-	-	HF
220	Llemma 7b	Llama Community	$0.80/M	-	-	HF
221	Qwen2.5 VL 72B Instruct	other	$0.80/M	73B	~$4/hr (2x A100)	HF
222	R1 Distill Llama 70B	MIT	$0.80/M	-	-	HF
223	Llama 3.1 Euryale 70B v2.2	CC BY	$0.85/M	-	-	HF
224	GLM 5.1	MIT	$0.97/M	754B	~$60+/hr (multi-node)	HF
225	GLM 5.2	Custom OSS	$1.00/M	-	-	-
226	Hermes 3 405B Instruct	Llama Community	$1.00/M	-	-	HF
227	Hermes 4 405B	Llama Community	$1.00/M	406B	~$30/hr (8x H100)	HF
228	Qwen-Max	Custom OSS	$1.04/M	-	-	-
229	Qwen3.6 Max Preview	Custom OSS	$1.04/M	-	-	-
230	Llama 3.1 Nemotron 70B Instruct	Llama Community	$1.20/M	71B	~$4/hr (2x A100)	HF
231	Qwen3.7 Max	Custom OSS	$1.25/M	-	-	-
232	Llama 3 Euryale 70B v2.1	CC BY	$1.48/M	-	-	HF
233	Mistral Medium 3.5	Custom OSS	$1.50/M	-	-	-
234	Jamba Large 1.7	other	$2.00/M	-	-	HF
235	Mistral Large	Custom OSS	$2.00/M	-	-	-
236	Mistral Large 2407	Custom OSS	$2.00/M	-	-	-
237	Mistral Large 2411	Custom OSS	$2.00/M	-	-	-
238	Mixtral 8x22B Instruct	Apache 2.0	$2.00/M	141B	~$12/hr (4x A100)	HF
239	Pixtral Large 2411	Custom OSS	$2.00/M	-	-	-
240	Command A	CC BY	$2.50/M	111B	~$12/hr (4x A100)	HF
241	Llama 3.1 70B Hanami x1	CC BY	$3.00/M	71B	~$4/hr (2x A100)	HF
242	Magnum v4 72B	Apache 2.0	$3.00/M	73B	~$4/hr (2x A100)	HF
243	Goliath 120B	Llama Community	$3.75/M	118B	~$12/hr (4x A100)	HF

Top 3 open-source LLMs

Gemma 3 12B (free) · Gemma License · cheapest API at $0.00/M input. Self-host estimate: -.

Gemma 3 27B (free) · Gemma License · cheapest API at $0.00/M input. Self-host estimate: ~$1.50/hr (1x A100).

Gemma 3 4B (free) · Gemma License · cheapest API at $0.00/M input. Self-host estimate: -.

Self-host vs API · when does it pay off?

Low volume

Under 1M tokens/day

Always pick API. A single GPU hour wipes out weeks of API spend at this volume.

Mid volume

10M to 100M tokens/day

Depends on model size. Small models (under 30B) are cheaper via API. Large models (200B+) favor dedicated GPUs.

High volume

1B+ tokens/day

Self-host wins, if utilization stays near 100%. Use reserved instances and bundle across workloads.

The price gap · cheapest vs most expensive

Cheapest

Gemma 3 12B (free)

$0.00/M

$ per 1M input tokens

Why the gap

Within OSS, the price range is driven by parameter count (bigger = more expensive to serve) and provider margins. Smaller dense models and heavily quantized serves sit at the bottom.

Most expensive

Goliath 120B

$3.75/M

$ per 1M input tokens

Frequently asked questions

The weights are publicly downloadable under some license (Apache 2.0, MIT, Llama Community, Qwen License, DeepSeek License, etc.). Not every "open" license is actually OSI-compliant · always read the license before commercial use.

Open source LLM pricing

Open source models · ranked by API price

Top 3 open-source LLMs

Self-host vs API · when does it pay off?

The price gap · cheapest vs most expensive

Frequently asked questions

See also

Licensing

Budget tiers

Compare