Llama 3.1 405B
The 405B member of the Meta Llama 3.1 collection of multilingual pretrained and instruction-tuned generative text models, intended for commercial and research use in multiple languages, including assistant-like chat.
text → text + code · made by Meta
Sold by 4 ways
| Seller | Lane | Rate |
|---|---|---|
| AWS Bedrock cloud | in-region | $1.2per Mtok in$1.2per Mtok outpricing.us-east-1.amazonaws.com · read 2026-08-24 |
| Snowflake Cortex cloud | standard | $1.2per Mtok in$1.2per Mtok outraw.githubusercontent.com · read 2026-09-16 |
| AWS Bedrock 0 ways | $5.32per Mtok in$16per Mtok out | |
| AWS Bedrock cloudnot on the seller's list today | standard | $5.32per Mtok in$16per Mtok outraw.githubusercontent.com · read 2026-08-27 |
| Oracle Cloud Infrastructure 0 ways | $10.68per Mtok in$10.68per Mtok out | |
| Oracle Cloud Infrastructure cloudnot on the seller's list today | standard | $10.68per Mtok in$10.68per Mtok outraw.githubusercontent.com · read 2026-08-26 |
Measured 10 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 1stof 134 | Arc ai2 — Epoch AI | Challenge score | 0.953 |
| 1stof 137 | Wino grande — Epoch AI | Accuracy | 89.2 |
| 3rdof 113 | Hella swag — Epoch AI | Overall accuracy | 0.892 |
| 4thof 88 | Bbh — Epoch AI | Average | 0.829 |
| 4thof 113 | Piqa — Epoch AI | Score | 85.9 |
| 15thof 217 | Mmlu — Epoch AI | EM | 0.844 |
| 19thof 115 | Trivia qa — Epoch AI | EM | 0.827 |
| 21stof 25 | SWE-Bench Pro (Public Dataset) | Resolve Rate | 11.18 |
| 240thof 302 | Artificial Analysis · Intelligence Index | Artificial Analysis Intelligence Index | 7 |
| 537thof 656 | Epoch capabilities index — Epoch AI | ECI Score | 128.75 |
About
Llama 3.1 405B — a text and code model from Meta, sold by 2 companies from $1.2 in and $1.2 out per million tokens, placed 1st of 137 on Wino grande — Epoch AI.
It takes text and returns text and code, with a context window of 131,072 tokens. It was published in September 2026. Its sellers say it can call a tool. The catalogue files it under code. Two companies sell it. The cheapest is $1.2 in and $1.2 out per million tokens at Snowflake Cortex. Beside the standard rate there is a regional lane. It has been measured on 10 boards, and stands best at 1st of 137 on Wino grande — Epoch AI.
Every current figure
- Maker
- Meta
- Register
- model
- Takes
- text
- Returns
- text + code
- Context
- 131,072 tokens
- Longest answer
- 8,192 tokens
- Published
- September 2026
- Parameters
- 405 billion · read from its own name
- Licence
- not read
- Sellers
- 2
- Maker's own price
- not read
- Price
- $1.2 in and $1.2 out per million tokens — Snowflake Cortex
- Boards
- 10
- Best place
- 1st of 137 — Wino grande — Epoch AI