Llama-3.1-70B-Instruct vs Qwen2.5-72B-Instruct
Specs, VRAM requirements and download trends — updated 30 July 2026.
meta-llama · A · Open weights
Params
70.6 B
Context
—
30d
203 K
llama3.1 licence · conditional
Specification comparison
Differences are highlighted; identical values are muted.
| Specification | Llama-3.1-70B-Instruct | Qwen2.5-72B-Instruct |
|---|---|---|
| Deployment | Open weights · also via API | Open weights · also via API |
| Parameters | 70.6 B | 72.7 B |
| Architecture | LlamaForCausalLM | Qwen2ForCausalLM |
| Context length | 131,072 | 32,768 |
| Licence | llama3.1 | other |
| Commercial use | Conditional | Unknown |
| Languages | 8 | 1 |
| Downloads 30d | 203,439 | 232,348 |
| Downloads all time | 24.5 M | 9.1 M |
| Quantizations on the Hub | SAFETENSORS | SAFETENSORS |
| Gated | Yes — accept licence | No |
| First seen on the Hub | 2024-07-16 | 2024-09-16 |
| API input price / 1M tokens | $0.40 | $0.36 |
| API output price / 1M tokens | $0.40 | $0.40 |
| Est. cost per 1M input + 1M output tokens | $0.80 | $0.76 |
| Cached input / 1M tokens | — | — |
| AA Intelligence Index | 6.6 | 7.7 |
| AA Coding Index | — | — |
| AA Agentic Index | — | — |
| Max output tokens | 16,384 | 16,384 |
| API providers | 2 | 1 |
| Tool calling | Yes | Yes |
| Reasoning mode | No | No |
API cost
Estimated cost of 1M input + 1M output tokens via OpenRouter
Llama-3.1-70B-Instruct
$0.80
Qwen2.5-72B-Instruct
$0.76
Qwen2.5-72B-Instruct is 1.1× cheaper at list price.
Download trend
Daily snapshots, last 89 days (28 Jul – 24 Oct)
1.0 M203 K
Llama-3.1-70B-Instruct
Qwen2.5-72B-Instruct
VRAM side by side
On RTX 4090 · 24 GB · 8K context unless noted
Llama-3.1-70B-Instruct · fp16
166.3 GB / 24 GB
Qwen2.5-72B-Instruct · fp16
171.4 GB / 24 GB