Provider Snapshot
11
79.92
0.55
Oct 17, 2025
Key Takeaways
11 together models are actively benchmarked with 10983 total measurements across 9726 benchmark runs.
llama-3.1-8b leads the fleet with 148.00 tokens/second, while llama-3.3-70b delivers 67.70 tok/s.
Performance varies by 118.6% across the together model lineup, indicating diverse optimization strategies for different use cases.
Median time to first token across the fleet is 0.55 ms, showing excellent responsiveness for interactive applications.
The together model fleet shows varied performance characteristics (46.0% variation coefficient), reflecting diverse model architectures.
Fastest Models
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| together | llama-3.1-8b | 148.00 | 5.14 | 231.00 | 0.22 |
| together | mistral-7b | 135.00 | 20.40 | 174.00 | 0.20 |
| together | llama-3.2-3b | 104.00 | 1.74 | 146.00 | 0.31 |
| together | qwen-2.5-7b | 103.00 | 4.86 | 145.00 | 0.21 |
| together | llama-3.1-70b | 87.20 | 1.14 | 158.00 | 0.41 |
| together | llama-3.3-70b | 67.70 | 1.98 | 143.00 | 0.69 |
All Models
Complete list of all together models tracked in the benchmark system. Click any model name to view detailed performance data.
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| together | llama-3.1-8b | 148.00 | 5.14 | 231.00 | 0.22 |
| together | llama-3.1-70b | 87.20 | 1.14 | 158.00 | 0.41 |
| together | llama-3.1-405b | 50.50 | 1.60 | 70.70 | 0.79 |
| together | llama-3.2-3b | 104.00 | 1.74 | 146.00 | 0.31 |
| together | mistral-7b | 135.00 | 20.40 | 174.00 | 0.20 |
| together | mixtral-8x7b | 62.00 | 2.25 | 111.00 | 0.64 |
| together | qwen-2.5-7b | 103.00 | 4.86 | 145.00 | 0.21 |
| together | qwen-2.5-72b | 47.50 | 1.13 | 71.30 | 0.40 |
| together | deepseek-v3 | 37.90 | 3.68 | 77.60 | 0.70 |
| together | deepseek-r1 | 36.30 | 1.01 | 66.60 | 1.43 |
| together | llama-3.3-70b | 67.70 | 1.98 | 143.00 | 0.69 |
Featured Models
Frequently Asked Questions
Based on recent tests, llama-3.1-8b shows the highest average throughput among tracked together models.
This provider summary aggregates 10983 individual prompts measured across 9726 monitoring runs over the past month.