Together AI API: models, pricing and limits
A practical guide to together, using data from the AI & LLM Index. The goal is a shortlist and a decision, not a tour of every option.
The Together AI lineup
Together AI has 14 models in the index. Inference and training cloud for open models. The lineup spans 3 categories, from multimodal to small open-weight.
The strongest model by average score is llama-3-1-turbo. The cheapest by input price is Mixtral Turbo 1.5B.
Pricing
Across the lineup, the average input price is $2.28 per million tokens and the average output price is $10.11. That average hides a wide range: individual models run from $0.31 to $4.21 on input.
Use the average as a starting point, then check the specific model you plan to deploy. Output price and cached input pricing can change the ranking for long or repetitive workloads.
Strengths and limits
14 of the 14 models are open weights. That gives you a self-hosting path for part of the lineup, while the proprietary models remain API-only.
The best benchmark result in the lineup is 76.8 on the index average. Compare that with the best model from other providers before treating Together AI as the default.
How to choose
- Start with the task, not the brand.
- Pick the cheapest model that clears your evaluation set.
- Keep a second provider as a fallback for outages and price changes.
- Check licence and data-handling terms for every model you shortlist.
Together AI models at a glance
| Model | Input / 1M | Context | Avg score |
|---|---|---|---|
| Llama 3.1 Turbo | $4.17 | 200,000 | 76.8 |
| Llama 3.1 Turbo Instruct | $4.21 | 32,768 | 76.7 |
| Llama 3.1 Turbo Chat | $3.95 | 200,000 | 76.1 |
| Llama 3.1 Turbo Base | $3.30 | 16,384 | 73.0 |
| Mixtral Turbo | $3.20 | 65,536 | 72.4 |
| Mixtral Turbo Instruct | $3.16 | 32,768 | 71.9 |
| Mixtral Turbo Chat | $3.02 | 32,768 | 71.7 |
| Mixtral Turbo Base | $2.53 | 32,768 | 69.7 |
| Llama 3.1 Turbo 8B | $1.16 | 32,768 | 58.8 |
| Llama 3.1 Turbo 7B | $0.98 | 8,192 | 56.4 |
Related
FAQ
Which Together AI model should I start with?
Start with llama-3-1-turbo for quality-sensitive work and Mixtral Turbo 1.5B for high-volume tasks.
What is the most important factor in together?
Fit to the task beats raw benchmark rank. Define the failure modes, pick the benchmarks that resemble them, then test a small shortlist on your own examples.
How many models should I evaluate?
Three is usually enough: a quality candidate, a value candidate and a fallback. More candidates add work without changing the decision.
How do I control cost?
Compare blended 3:1 price, watch output tokens, use caching where available and route simple requests to a cheaper model.