DeepSeek R2 ships 32B model, rivals GPT-5 on reasoning at lower cost

作者 PulseAugur 编辑部 · [1 个来源] · 2026-04-29 08:16

DeepSeek has released its R2 model, a 32 billion parameter dense transformer. This new model achieves 92.7% accuracy on the AIME 2025 benchmark and can operate on a single RTX 4090 graphics card. The R2 model is also significantly more cost-effective, costing approximately 70% less than GPT-5 for reasoning tasks, and is available under an MIT license for self-hosting. AI

影响 Offers a cost-effective, high-performance alternative for reasoning tasks, potentially impacting enterprise adoption and research.

排序理由 New model release from a significant AI lab with strong benchmark performance.

在 Mastodon — fosstodon.org 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

报道来源 [1]

Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] · 2026-04-29 08:16

DeepSeek R2 just shipped: 32B dense transformer, 92.7% on AIME 2025, runs on a single RTX 4090. It costs ~70% less than GPT-5 for reasoning tasks. MIT licensed,

DeepSeek R2 just shipped: 32B dense transformer, 92.7% on AIME 2025, runs on a single RTX 4090. It costs ~70% less than GPT-5 for reasoning tasks. MIT licensed, self-hostable. Full developer guide — benchmarks, pricing, API access: https:// crazyrouter.com/en/blog/deepse ek-r2-re…

报道来源 [1]

DeepSeek R2 just shipped: 32B dense transformer, 92.7% on AIME 2025, runs on a single RTX 4090. It costs ~70% less than GPT-5 for reasoning tasks. MIT licensed,

相关实体

相关话题