Meta · October 2026
Llama API pricing, every model
16 Llama models, from $0.05 to $4 per million input tokens. Meta’s Llama models, served by many hosts at different prices.
Prices from OpenRouter’s models API, updated October 29, 2026
Llama 3.1 8B Instruct
$0.05 in · $0.08 outAbout $3.21 a month for a 1,000-chats-a-day support bot.
Muse Spark Open
$1.25 in · $4.25 outReleased 2026-10-15, 1.05M tokens of context.
Muse Spark 1.3 Contributor
−98%Cached input costs $0.002 instead of $0.10 per 1M tokens.
All Llama models and prices
| Model | Released | Input /1M | Output /1M | Cache read /1M | Context | Chatbot, per month* |
|---|---|---|---|---|---|---|
| Muse Spark Open | 2026-10-15 | $1.25 | $4.25 | $0.15 | 1.05M | $107 |
| Muse Wonder | 2026-10-12 | $4 | $20 | $0.20 | 1.05M | $420 |
| Muse Spark 1.3 | 2026-09-02 | $1.25 | $4.25 | $0.15 | 1.05M | $107 |
| Muse Spark 1.3 Contributor | 2026-09-02 | $0.10 | $0.20 | $0.002 | 1.05M | $6.90 |
| Muse Spark 1.2 Contributor | 2026-08-21 | $0.10 | $0.20 | $0.002 | 1.05M | $6.90 |
| Muse Glimmer 30B | 2026-08-09 | $0.30 | $1.20 | $0.04 | 131K | $27.90 |
| Muse Spark 1.2 | 2026-08-05 | $1.25 | $4.25 | $0.15 | 1.05M | $107 |
| Muse Spark 1.1 | 2026-07-16 | $1.25 | $4.25 | $0.15 | 1.05M | $107 |
| Llama Guard 4 12B | 2025-04-30 | $0.18 | $0.18 | – | 164K | $10.26 |
| Llama 4 Maverick | 2025-04-05 | $0.188 | $0.652 | $0.05 | 524K | $16.27 |
| Llama 4 Scout | 2025-04-05 | $0.10 | $0.30 | – | 1.31M | $8.10 |
| Llama 3.3 70B Instruct | 2024-12-06 | $0.10 | $0.32 | – | 131K | $8.34 |
| Llama 3.2 1B Instruct | 2024-09-25 | $0.027 | $0.201 | – | 60K | $3.63 |
| Llama 3.2 3B Instruct | 2024-09-25 | $0.05 | $0.33 | – | 131K | $6.21 |
| Llama 3.1 70B Instruct | 2024-07-23 | $0.40 | $0.40 | – | 131K | $22.80 |
| Llama 3.1 8B Instruct | 2024-07-23 | $0.05 | $0.08 | $0.025 | 131K | $3.21 |
* Support chatbot: 1,000 chats a day, 1,500 tokens in, 400 out, 30 days, no caching. Prices in US dollars per 1M tokens.
What three real workloads cost a month
| Model | Support chatbot | Daily summaries | Coding agent |
|---|---|---|---|
| Llama 3.1 8B Instruct | $3.21 | $52.20 | $23.52 |
| Llama 3.2 1B Instruct | $3.63 | $42.39 | $24.26 |
| Llama 3.2 3B Instruct | $6.21 | $74.70 | $43.92 |
| Muse Spark 1.3 Contributor | $6.90 | $108 | $20.35 |
| Muse Spark 1.2 Contributor | $6.90 | $108 | $20.35 |
| Llama 4 Scout | $8.10 | $117 | $79.20 |
| Llama 3.3 70B Instruct | $8.34 | $119 | $79.68 |
| Llama Guard 4 12B | $10.26 | $178 | $134 |
Support chatbot: 1,000 chats a day, 1,500 tokens in, 400 out · Daily summaries: 10,000 a day, 3,000 tokens in, 300 out · Coding agent: 400 calls a day, 60,000 tokens in (80% cached), 2,000 out. Cached share billed at the cache-read price where the model has one.
Compare with other providers
Llama pricing questions
How much does the Llama API cost?
Meta’s models range from $0.05 per 1M input tokens and $0.08 per 1M output tokens (Llama 3.1 8B Instruct) to $4 and $20 (Muse Wonder). The newest, Muse Spark Open, costs $1.25 in and $4.25 out.
Which Llama model is the cheapest?
Llama 3.1 8B Instruct, at $0.05 per 1M input tokens. A support chatbot doing 1,000 chats a day would cost about $3.21 a month on it.
Does Meta discount cached prompts?
Yes. On Muse Spark 1.3 Contributor, cached input costs $0.002 per 1M tokens instead of $0.10, 98% less. Long, repeated system prompts and agent loops benefit most.
How do I track my Llama spend per product?
Route it through OpenRouter with one API key per product and connect OpenRouter to Glow: you get spend by model, API key and product every day, with cache hit rate. Track AI spend.
See your real Llama bill.
Glow shows what you actually spend by model, API key and product, every day, and which cheaper model would do the same job.
- Read-only keys
- Never shown again
- 2FA + Face ID
- 7-day trial