DeepSeek V4 闪存 0731

DeepSeek V4 Flash 0731 DeepSeek · 2026 年 7 月 31 日 · 3 个推理变体 论文 ↗ 模型 ↗ 在最大努力下，DeepSeek V4 Flash 0731 在每任务 0.02 美元的 ARC-AGI-1 半私有上得分 89.0%，在每任务 0.04 美元的 ARC-AGI-2 半私有上得分 61.4%。 ARC-AGI 2 排行榜 DeepSeek V4 Flash 0731 验证分数 变体 ARC-AGI-1 ARC-AGI-2 ARC-AGI-3 最大 89.0% 61.4% — 高 87.0% 56.0% — 低 84.0% 46.0% — 任务和环境 每个基准测试中每个推理级别的通过/失败。 ARC-AGI-2 公共评估 120 任务 任务最大高低 0934a4d8 ✗ ✓ ✗ 135a2760 ✓ ✓ ✓ 136b0064 ✓ ✓ ✓ 13e47133 ✗ ✗ ✗ 142ca369 ✗ ✓ ✗ 16b78196 ✗ ✗ ✗ 16de56c4 ✓ ✓ ✓ 1818057f ✓ ✓ ✓ 195c6913 ✓ ✗ ✗ 1ae2feb7 ✓ ✓ ✓ 20270e3b ✓ ✓ ✗ 20a9e565 ✗ ✗ ✗ 21897d95 ✗ ✗ ✗ 221dfab4 ✓ ✓ ✗ 247ef758 ✓ ✓ ✓ 269e22fb ✗ ✗ ✗ 271d71e2 ✗ ✓ ✗ 28a6681f ✓ ✗ ✗ 291dc1e1 ✓ ✓ ✗ 2b83f449 ✗ ✗ ✗ 2ba387bc ✓ ✓ ✓ 2c181942 ✓ ✓ ✗ 2d0172a1 ✗ ✗ ✗ 31f7f899 ✓ ✓ ✓ 332f06d7 ✓ ✗ ✗ 35ab12c3 ✓ ✓ ✓ 36a08778 ✓ ✓ ✓ 38007db0 ✓ ✓ ✓ 3a25b0d8 ✗ ✗ ✗ 3dc255db ✗ ✗ ✗ 3e6067c3 ✓ ✗ ✗ 409aa875 ✓ ✓ ✗ 446ef5d2 ✗ ✗ ✗ 45a5af55 ✓ ✓ ✓ 4a21e3da ✓ ✗ ✓ 4c3d4a41 ✗ ✓ ✓ 4c416de3 ✗ ✗ ✗ 4c7dc4dd ✗ ✗ ✗ 4e34c42c ✗ ✗ ✗ 53fb4810 ✓ ✓ ✓ 5545f144 ✗ ✗ ✗ 581f7754 ✗ ✗ ✗ 58490d8a ✓ ✓ ✓ 58f5dbd5 ✓ ✓ ✓ 5961cc34 ✓ ✓ ✓ 5dbc8537 ✓ ✗ ✗ 62593bfd ✗ ✗ ✗ 64efde09 ✗ ✓ ✗ 65b59efc ✓ ✓ ✓ 67e490f4 ✓ ✓ ✓ 6e453dd6 ✓ ✓ ✓ 6e4f6532 ✓ ✗ ✗ 6ffbe589 ✗ ✗ ✗ 71e489b6 ✓ ✓ ✗ 7491f3cf ✓ ✗ ✓ 7666fa5d ✓ ✓ ✗ 78332cb0 ✗ ✗ ✗ 7b0280bc ✓ ✗ ✗ 7b3084d4 ✓ ✓ ✗ 7b5033c1 ✗ ✗ ✓ 7b80bb43 ✗ ✗ ✗ 7c66cb00 ✗ ✗ ✓ 7ed72f31 ✓ ✓ ✗ 800d221b ✗ ✗ ✗ 80a900e0 ✓ ✓ ✓ 8698868d ✗ ✗ ✗ 88bcf3b4 ✗ ✗ ✗ 88e364bc ✗ ✗ ✗ 89565ca0 ✓ ✗ ✓ 898e7135 ✓ ✓ ✓ 8b7bacbf ✗ ✗ ✗ 8b9c3697 ✓ ✗ ✗ 8e5c0c38 ✓ ✗ ✓ 8f215267 ✓ ✓ ✓ 8f3a5a89 ✓ ✓ ✓ 9385bd28 ✓ ✓ ✗ 97d7923e ✓ ✗ ✓ 981571dc ✓ ✓ ✓ 9aaea919 ✓ ✓ ✓ 9bbf930d ✗ ✗ ✗ a251c730 ✓ ✓ ✓ a25697e4 ✗ ✗ ✗ a32d8b75 ✗ ✗ ✗ a395ee82 ✓ ✓ ✓ a47bf94d ✗ ✓ ✓ a6f40cea ✓ ✓ ✓ aa4ec2a5 ✓ ✓ ✓ abc82100 ✗ ✗ ✗ b0039139 ✓ ✗ ✓ b10624e5 ✓ ✓ ✓ b5ca7ac4 ✓ ✓ ✓ b6f77b65 ✓ ✓ ✗ b99e7126 ✓ ✓ ✓ b9e38dc0 ✗ ✗ ✗ bf45cf4b ✓ ✓ ✓ c4d067a0 ✓ ✓ ✓ c7f57c3e ✓ ✗ ✗ cb2d8a2c ✓ ✓ ✗ cbebaa4b ✓ ✓ ✗ d35bdbdc ✗ ✗ ✗ d59b0160 ✓ ✓ ✓ d8e07eb2 ✓ ✓ ✓ da515329 ✗ ✗ ✗ db0c5428 ✓ ✓ ✓ db695cfb ✓ ✓ ✓ dbff022c ✓ ✓ ✗ dd6b8c4b ✗ ✗ ✗ de809cff ✗ ✗ ✗ dfadab01 ✗ ✓ ✗ e12f9a14 ✗ ✗ ✗ e3721c99 ✗ ✗ ✗ e376de54 ✓ ✓ ✗ e8686506 ✓ ✓ ✓ e87109e9 ✓ ✓ ✗ edb79dae ✗ ✓ ✓ eee78d87 ✓ ✗ ✗ f560132c ✗ ✗ ✗ f931b4a8 ✓ ✓ ✓ faa9f03d ✗ ✗ ✗ fc7cae8d ✗ ✓ ✗ ARC-AGI-1 公共评估 400 任务 任务最大高低 00576224 ✓ ✓ ✓ 009d5c81 ✓ ✓ ✓ 00dbd492 ✓ ✓ ✓ 03560426 ✓ ✓ ✓ 05a7bcf2 ✓ ✓ ✓ 0607ce86 ✓ ✓ ✓ 0692e18c ✓ ✓ ✓ 070dd51e ✓ ✓ ✓ 08573cc6 ✓ ✓ ✓ 0934a4d8 ✓ ✓ ✓ 09c534e7 ✓ ✓ ✗ 0a1d4ef5 ✓ ✓ ✓ 0a2355a6 ✓ ✓ ✓ 0b17323b ✓ ✓ ✓ 0bb8deee ✓ ✓ ✓ 0becf7df ✓ ✓ ✓ 0c786b71 ✓ ✓ ✓ 0c9aba6e ✓ ✓ ✓ 0d87d2a6 ✗ ✗ ✗ 0e671a1a ✓ ✓ ✓ 0f63c0b9 ✓ ✓ ✓ 103eff5b ✓ ✓ ✓ 11e1fe23 ✓ ✓ ✓ 12422b43 ✓ ✓ ✓ 12997ef3 ✓ ✓ ✓ 12eac192 ✓ ✓ ✓ 136b0064 ✓ ✓ ✓ 13713586 ✓ ✓ ✓ 137f0df0 ✓ ✓ ✓ 140c817e ✓ ✓ ✓ 14754a24 ✓ ✗ ✗ 15113be4 ✓ ✓ ✓ 15663ba9 ✓ ✓ ✓ 15696249 ✓ ✓ ✓ 16b78196 ✓ ✗ ✗ 17b80ad2 ✓ ✓ ✓ 17cae0c1 ✓ ✓ ✓ 18419cfa ✓ ✓ ✓ 184a9768 ✓ ✓ ✓ 195ba7dc ✓ ✓ ✓ 1990f7a8 ✓ ✓ ✓ 19bb5feb ✓ ✓ ✓ 1a2e2828 ✓ ✓ ✓ 1a6449f1 ✓ ✓ ✓ 1acc24af ✗ ✗ ✗ 1c02dbbe ✓ ✓ ✓ 1c0d0a4b ✓ ✓ ✓ 1c56ad9f ✓ ✓ ✓ 1d0a4b61 ✓ ✓ ✓ 1d398264 ✓ ✓ ✓ 1da012fc ✓ ✓ ✓ 1e81d6f9 ✓ ✓ ✓ 1e97544e ✓ ✓ ✓ 2037f2c7 ✓ ✓ ✓ 2072ab