Llama Guard 4 12B is 9.5× cheaper per million tokens (blended). Qwen3.8 27B has the larger context window (1,000,000 vs 163,840 tokens, 6.1×).
Llama Guard 4 12B vs Qwen3.8 27B
Live catalog fields. Best-in-row is highlighted.
Llama Guard 4 12Bopen
Meta · meta-llama__llama-guard-4-12b
Qwen3.8 27Bopen
Qwen · qwen__qwen3.8-27b
Identity
| Field | meta-llama__llama-guard-4-12b | qwen__qwen3.8-27b |
|---|---|---|
| Provider | Meta | Qwen |
| Slug | meta-llama__llama-guard-4-12b | qwen__qwen3.8-27b |
| Status | live | live |
| Open weights | Yes | Yes |
| License | — | — |
| HuggingFace | meta-llama/Llama-Guard-4-12B | Qwen/Qwen3.8-27B |
Quality
| Field | meta-llama__llama-guard-4-12b | qwen__qwen3.8-27b |
|---|---|---|
| Quality index (0–100) | — | 87.4 |
Cost
| Field | meta-llama__llama-guard-4-12b | qwen__qwen3.8-27b |
|---|---|---|
| Input $ / 1M tokens | $0.18 | $0.42 |
| Cached input $ / 1M | — | $0.09 |
| Output $ / 1M tokens | $0.18 | $3.00 |
Context
| Field | meta-llama__llama-guard-4-12b | qwen__qwen3.8-27b |
|---|---|---|
| Max context tokens | 163,840 | 1,000,000 |
| Max output tokens | 16,384 | 131,072 |
Modalities
| Field | meta-llama__llama-guard-4-12b | qwen__qwen3.8-27b |
|---|---|---|
| Input | image, text | text, image, video |
| Output | text | text |