Llama API pricing
Meta's earlier open-weight models, priced by third-party hosts. Meta has not released a new Llama model in 2026.
Llama 4 Maverick
PreviousPrevious generation. The current replacement is Muse Spark 1.3 at $1.25 / $4.25 per 1M tokens.
What it costs in practice
Llama 4 Maverick at list prices, no batch discount.
| 1,000 chat replies1,500 tokens in and 300 out each | $0.48 |
|---|---|
| Summarising 100 long reports20,000 tokens in and 800 out each | $0.43 |
| A month of a coding agent400 tasks, 85% of input cached | $146 |
The coding-agent example assumes 85% of input is served from cache, typical for agents that resend the same context each step. Model your own workload
Details
llama-4-maverickCheapest host listed on OpenRouter. Together AI and Groq dropped it from their standard serverless tiers in 2026.
Checked 27 Sep 2026 against OpenRouter listing and Groq deprecations .
Other Llama versions
Several hosts dropped Llama 4 from their standard tiers in 2026.
| Model | Input / 1M | Output / 1M | Cached input | Context | Released |
|---|---|---|---|---|---|
Llama 3.3 70BPrevious Open weights · via OpenRouter |
$0.10 | $0.32 | — | 128K | 6 Dec 2024 |
Llama 4 ScoutPrevious Open weights · via OpenRouter |
$0.10 | $0.30 | — | — | 5 Apr 2025 |
Version details
Llama 3.3 70B
Previous$0.10 / $0.32 per 1M tokens128K context
Cheapest host listed on OpenRouter. Together AI charges $1.04 / $1.04; Groq limited it to enterprise customers from 16 August 2026.
Newer option: Muse Glimmer 30B at $0.30 / $1.20. Checked 27 Sep 2026 against OpenRouter listing and Together AI pricing .
Llama 4 Scout
Previous$0.10 / $0.30 per 1M tokens
Cheapest host listed on OpenRouter. Groq dropped it for non-enterprise users on 17 July 2026.
Newer option: Muse Glimmer 30B at $0.30 / $1.20. Checked 27 Sep 2026 against OpenRouter listing .
Compare Llama 4 Maverick
No ready-made comparisons for this family yet.
Price history
All changesNo price changes recorded since we began tracking in June 2026.
Meta pricing rules
- Muse Spark has a Contributor tier at $0.10 / $0.20 per 1M tokens if Meta may train on your traffic.
- Open-weight Meta models (Muse Glimmer, Llama) are priced by third-party hosts.
Questions
How much does Llama 4 Maverick cost?
As of 27 Sep 2026, Llama 4 Maverick costs $0.1875 per million input tokens and $0.65 per million output tokens. Prices checked 27 Sep 2026.
What is the cheapest Llama model?
Llama 4 Scout, at $0.10 / $0.30 per million input / output tokens.
What is Llama 4 Maverick's context window?
1M tokens.