Claude 3 Haiku is 2.3× cheaper per million tokens (blended). Qwen3.8 27B has the larger context window (1,000,000 vs 200,000 tokens, 5.0×). Qwen3.8 27B is open-weights; the other is not.
Claude 3 Haiku vs Qwen3.8 27B
Live catalog fields. Best-in-row is highlighted.
Claude 3 Haiku
Anthropic · anthropic__claude-3-haiku
Qwen3.8 27Bopen
Qwen · qwen__qwen3.8-27b
Identity
| Field | anthropic__claude-3-haiku | qwen__qwen3.8-27b |
|---|---|---|
| Provider | Anthropic | Qwen |
| Slug | anthropic__claude-3-haiku | qwen__qwen3.8-27b |
| Status | live | live |
| Open weights | No | Yes |
| License | — | — |
| HuggingFace | — | Qwen/Qwen3.8-27B |
Quality
| Field | anthropic__claude-3-haiku | qwen__qwen3.8-27b |
|---|---|---|
| Quality index (0–100) | — | 89.0 |
Cost
| Field | anthropic__claude-3-haiku | qwen__qwen3.8-27b |
|---|---|---|
| Input $ / 1M tokens | $0.25 | $0.42 |
| Cached input $ / 1M | $0.03 | $0.09 |
| Output $ / 1M tokens | $1.25 | $3.00 |
Context
| Field | anthropic__claude-3-haiku | qwen__qwen3.8-27b |
|---|---|---|
| Max context tokens | 200,000 | 1,000,000 |
| Max output tokens | 4,096 | 131,072 |
Modalities
| Field | anthropic__claude-3-haiku | qwen__qwen3.8-27b |
|---|---|---|
| Input | text, image | text, image, video |
| Output | text | text |