Model family · 8 versions tracked
Llama: every version compared
Meta: Muse Spark 1.1 currently leads the Llama family with an Artificial Analysis Intelligence Index of 53.2.
- Meta: Muse Spark 1.1 - 53.2 intelligence · $2.00/1M
- Meta: Muse Spark 1.2 - - intelligence · $2.00/1M
- Meta: Llama Guard 4 12B - - intelligence · $0.18/1M
- Meta: Llama 3.3 70B Instruct - - intelligence · $0.15/1M
- Meta: Llama 3.2 1B Instruct - - intelligence · $0.07/1M
- Meta: Llama 3.2 3B Instruct - - intelligence · $0.12/1M
- Meta: Llama 3.1 70B Instruct - - intelligence · $0.40/1M
- Meta: Llama 3.1 8B Instruct - - intelligence · $0.06/1M
| # | Model | Intelligence index | Price/1M blended |
|---|---|---|---|
| 01 | Meta: Muse Spark 1.1 | 53.2 | $2.00 |
| 02 | Meta: Muse Spark 1.2 | - | $2.00 |
| 03 | Meta: Llama Guard 4 12B | - | $0.18 |
| 04 | Meta: Llama 3.3 70B Instruct | - | $0.15 |
| 05 | Meta: Llama 3.2 1B Instruct | - | $0.07 |
| 06 | Meta: Llama 3.2 3B Instruct | - | $0.12 |
| 07 | Meta: Llama 3.1 70B Instruct | - | $0.40 |
| 08 | Meta: Llama 3.1 8B Instruct | - | $0.06 |
Llama real usage share, over time
| # | Model | Share of usage |
|---|---|---|
| 01 | Llama usage share | 0.2% |
Daily tracking began Jul 5- this becomes a live chart once there's a third day to plot.
Ranked by current quality, not release date - this pipeline's data collection started recently, so "first tracked" timestamps don't reflect true historical release dates yet. The usage-share chart above is real, unaffected by that limitation.
Frequently asked questions
What is the best Llama model?
Meta: Muse Spark 1.1 currently leads the Llama family with an Artificial Analysis Intelligence Index of 53.2.
How much real usage does Llama get?
Llama models account for 0.2% of real OpenRouter token volume as of the latest data.