Llama-3.1-8B-Instruct vs Qwen3-4B-unsloth-bnb-4bit
Specs, VRAM requirements and download trends — updated 2 September 2026.
meta-llama · A · Open weights
Params
8.0 B
Context
—
30d
4.5 M
llama3.1 licence · conditional
unsloth · B · Open weights
Params
4.1 B
Context
40,960
30d
188 K
apache-2.0 · commercial OK
Specification comparison
Differences are highlighted; identical values are muted.
| Specification | Llama-3.1-8B-Instruct | Qwen3-4B-unsloth-bnb-4bit |
|---|---|---|
| Deployment | Open weights · also via API | Open weights |
| Parameters | 8.0 B | 4.1 B |
| Architecture | LlamaForCausalLM | Qwen3ForCausalLM |
| Context length | 131,072 | 40,960 |
| Licence | llama3.1 | apache-2.0 |
| Commercial use | Conditional | Allowed |
| Languages | 8 | 1 |
| Downloads 30d | 4,504,499 | 187,861 |
| Downloads all time | 193.6 M | 1.5 M |
| Quantizations on the Hub | SAFETENSORS | SAFETENSORS |
| Gated | Yes — accept licence | No |
| First seen on the Hub | 2024-07-18 | 2025-04-28 |
| API input price / 1M tokens | $0.02 | not on OpenRouter |
| API output price / 1M tokens | $0.04 | not on OpenRouter |
| Est. cost per 1M input + 1M output tokens | $0.06 | not on OpenRouter |
| Cached input / 1M tokens | — | not on OpenRouter |
| AA Intelligence Index | — | not on OpenRouter |
| AA Coding Index | 5.4 | not on OpenRouter |
| AA Agentic Index | — | not on OpenRouter |
| Max output tokens | 16,384 | not on OpenRouter |
| API providers | 4 | not on OpenRouter |
| Tool calling | Yes | not on OpenRouter |
| Reasoning mode | No | not on OpenRouter |
API cost
Estimated cost of 1M input + 1M output tokens via OpenRouter
Llama-3.1-8B-Instruct
$0.06
Download trend
Daily snapshots, last 90 days (13 Aug – 10 Nov)
7.6 M176 K
Llama-3.1-8B-Instruct
Qwen3-4B-unsloth-bnb-4bit
VRAM side by side
On RTX 4090 · 24 GB · 8K context unless noted
Qwen3-4B-unsloth-bnb-4bit · fp16
5.0 GB / 24 GB
Llama-3.1-8B-Instruct · fp16
19.4 GB / 24 GB