OptimToken › Models by maker › Meta

Meta API pricing

As of 2026-10-05, the OptimToken catalogue holds 12 Meta models. By price tier, 3 are Mid-tier and 9 Budget. The cheapest is Llama 3.1 8B Instruct at $0.05 per 1M input tokens and $0.08 per 1M output tokens. The most expensive is Muse Spark 1.3 at $1.25 input and $4.25 output, 47x the cheapest on blended price (30% input, 70% output), the same blended price as 2 others. These are the prices we quote: a model sold by several hosts has other rates, listed on its own page.

Meta models and prices

Meta models, cheapest first on blended price (30% input, 70% output). USD per 1M tokens, data snapshot 2026-10-05.
Model	Input	Output	Batch (in / out)	Cache read	Context	Licence
Llama 3.1 8B Instruct	$0.05	$0.08	not published	$0.03	131K	Open weights (Llama 3.x)
Muse Spark 1.2 Contributor	$0.10	$0.20	not published	$0.0020	1M	Unknown
Muse Spark 1.3 Contributor	$0.10	$0.20	not published	$0.0020	1M	Unknown
Llama Guard 4 12B	$0.18	$0.18	not published	not published	164K	Open weights (Llama 4)
Llama 4 Scout	$0.10	$0.30	not published	not published	1M	Open weights (Llama 4)
Llama 3.3 70B Instruct	$0.10	$0.32	not published	not published	131K	Open weights (Llama 3.x)
Llama 3.1 70B Instruct	$0.40	$0.40	not published	not published	131K	Open weights (Llama 3.x)
Llama 4 Maverick	$0.19	$0.65	not published	not published	1M	Open weights (Llama 4)
Muse Glimmer 30B	$0.35	$1.50	not published	$0.04	131K	Unknown
Muse Spark 1.1	$1.25	$4.25	not published	$0.15	1M	Unknown
Muse Spark 1.2	$1.25	$4.25	not published	$0.15	1M	Unknown
Muse Spark 1.3	$1.25	$4.25	not published	$0.15	1M	Unknown

Published discounts: a prompt-cache read rate for 7 of the 12 and a batch rate for none. The most recent by release date are Muse Spark 1.3 and Muse Spark 1.3 Contributor (2026-09). Arena ELO scores (as of 2026-10-02) exist for 5 of the 12; the highest is Llama 4 Maverick at 1327.

Licences

Of the 12, 6 are open weights, which is not the same as open source: the weights are published, but under the Llama 3.x and Llama 4 licences, which restrict some uses. Read them before self-hosting commercially. No licence is recorded for the other 6, so this page does not say whether they can be self-hosted. Licences are as recorded in this catalogue: check the maker's own terms before relying on one.

Price changes since tracking began

Our daily price archive began on 2026-08-15 and holds a record for all 12 models. The price we quote has changed for 5 of them, 21 times in all, 15 of those changes flagged as a change of quoted host rather than a repricing. The most recent was to Llama 3.3 70B Instruct on 2026-10-05: -38.9% on the blended price (30% input, 70% output) (quoted host changed, not a repricing). Since tracking began, 6 of them joined the catalogue, most recently Muse Spark 1.3 on 2026-09-03. Last checked against the 2026-10-05 snapshot.

Each row is a model's most recent change. Its own page has the full history.

Model	Date	Old price (in / out)	New price (in / out)	Blended change	Changes in all
Llama 3.3 70B Instruct	2026-10-05	$0.22 / $0.50	$0.10 / $0.32	-38.9% (quoted host changed, not a repricing)	4
Muse Glimmer 30B	2026-09-30	$0.30 / $1.20	$0.35 / $1.50	+24.2% (quoted host changed, not a repricing)	9
Llama 3.1 70B Instruct	2026-09-22	$0.72 / $0.72	$0.40 / $0.40	-44.4% (quoted host changed, not a repricing)	2
Llama 4 Maverick	2026-09-22	$0.20 / $0.80	$0.19 / $0.65	-17.3% (quoted host changed, not a repricing)	4
Llama 4 Scout	2026-09-01	$0.11 / $0.34	$0.10 / $0.30	-11.4% (quoted host changed, not a repricing)	2

Head-to-head comparisons

Models from Meta appear on 1 ready-made side-by-side page.

Llama 4 Maverick vs DeepSeek V3.2

Other makers

Models by maker: every model in the catalogue, including makers with fewer than 3.

These are list prices.

Your real cost also depends on caching and batching, negotiated discounts, marketplace agreements, and the contracts you already have. OptimNow audits exactly that.

Get a FinOps review