AI BENCHY Compare
Anthropic: Claude Opus 4.6 vs DeepSeek: DeepSeek V3.2
Benchmark dihasilkan dari suite pengujian AI BENCHY pada: 2026-03-03
| Metrik | Anthropic: Claude Opus 4.6 medium Rilis: 2026-02-05 | DeepSeek: DeepSeek V3.2 none Rilis: 2025-12-01 |
|---|---|---|
| Peringkat | #21 | #30 |
| Skor Rata-rata | 6.07 | 4.70 |
| Konsistensi | 8.80 | 8.19 |
| Biaya per hasil | 10.992 | 0.132 |
| Total Biaya | $0.880 | $0.007 |
| Tes benar | 8/14 | 5/14 |
| Tingkat lulus per percobaan | 61.9% | 47.6% |
| Tes tidak stabil | 2 | 3 |
| Token output | 19,576 | 4,869 |
| Token penalaran | 11,081 | 0 |
Rincian Kategori
| Trik anti-AI | Skor | Konsistensi | Tingkat lulus per percobaan | Tes tidak stabil | Tes benar | Token output | Token penalaran |
|---|---|---|---|---|---|---|---|
| Anthropic: Claude Opus 4.6 | 4.00 | 4.41 | 55.6% | 2 | 897 | 1,000 | |
| DeepSeek: DeepSeek V3.2 | 1.00 | 9.68 | 0.0% | 0 | 1,411 | 0 |
| Parsing dan ekstraksi data | Skor | Konsistensi | Tingkat lulus per percobaan | Tes tidak stabil | Tes benar | Token output | Token penalaran |
|---|---|---|---|---|---|---|---|
| Anthropic: Claude Opus 4.6 | 10.00 | 10.00 | 100.0% | 0 | 668 | 763 | |
| DeepSeek: DeepSeek V3.2 | 5.38 | 5.81 | 66.7% | 1 | 1,710 | 0 |
| Spesifik domain | Skor | Konsistensi | Tingkat lulus per percobaan | Tes tidak stabil | Tes benar | Token output | Token penalaran |
|---|---|---|---|---|---|---|---|
| Anthropic: Claude Opus 4.6 | 1.00 | 10.00 | 0.0% | 0 | 16,328 | 7,928 | |
| DeepSeek: DeepSeek V3.2 | 1.00 | 7.21 | 22.2% | 1 | 24 | 0 |
| Kepatuhan instruksi | Skor | Konsistensi | Tingkat lulus per percobaan | Tes tidak stabil | Tes benar | Token output | Token penalaran |
|---|---|---|---|---|---|---|---|
| Anthropic: Claude Opus 4.6 | 9.50 | 9.99 | 100.0% | 0 | 266 | 468 | |
| DeepSeek: DeepSeek V3.2 | 8.00 | 9.99 | 50.0% | 0 | 66 | 0 |
| Puzzle Solving | Skor | Konsistensi | Tingkat lulus per percobaan | Tes tidak stabil | Tes benar | Token output | Token penalaran |
|---|---|---|---|---|---|---|---|
| Anthropic: Claude Opus 4.6 | 7.00 | 10.00 | 66.7% | 0 | 556 | 593 | |
| DeepSeek: DeepSeek V3.2 | 7.67 | 7.49 | 88.9% | 1 | 1,136 | 0 |
| Pemanggilan alat | Skor | Konsistensi | Tingkat lulus per percobaan | Tes tidak stabil | Tes benar | Token output | Token penalaran |
|---|---|---|---|---|---|---|---|
| Anthropic: Claude Opus 4.6 | 10.00 | 10.00 | 100.0% | 0 | 861 | 329 | |
| DeepSeek: DeepSeek V3.2 | 10.00 | 10.00 | 100.0% | 0 | 522 | 0 |
Perbandingan Cepat
Ganti Pasangan Perbandingan
Claude Opus 4.6mediumvsGLM 5noneClaude Opus 4.6mediumvsQwen3.5 Plus 2026-02-15noneDeepSeek V3.2nonevsGLM 4.7 FlashmediumClaude Opus 4.6mediumvsQwen3.5-122B-A10BnoneDeepSeek V3.2nonevsMiniMax M2.5mediumDeepSeek V3.2nonevsgpt-oss-120bmediumTersedia gratisDeepSeek V3.2nonevsQwen3.5-FlashmediumClaude Opus 4.6mediumvsGemini 3 Flash PreviewnoneDeepSeek V3.2nonevsGPT-5 NanomediumDeepSeek V3.2nonevsQwen3 Coder NextmediumClaude Opus 4.6mediumvsGPT-4o-mininoneDeepSeek V3.2nonevsQwen3.5-35B-A3Bmedium