Tag
A team slashed AI workflow costs from $62,000 to $7,800 per month by using Claude Opus 4.8 for orchestration and Kimi K2.6 Agent Swarm for execution, with a detailed 15-prompt system.
Moonshot AI founder Yang Zhilin released a 40-minute video detailing the training process of the Kimi K2 model, which cost only $4.6 million. In an 8-model real-time programming competition, Kimi K2 took first place, defeating GPT-5.5 and others, demonstrating how a small team can overturn the traditional compute-stacking paradigm through architecture optimization.
DeepSeek released V4 Pro and V4 Flash under MIT license on April 24, 2026. In benchmarks against Claude Opus 4.7 and Kimi K2.6, V4 Pro scored 77/100 at $2.25, placing between Opus 4.7 (91) and Kimi K2.6 (68), while V4 Flash scored 60/100 at $0.02, the cheapest in the comparison, with a 75% discount on V4 Pro through May 31.
Kimi K2, trained for $4.6 million, outperforms GPT-5 and Claude Opus 4.7 on coding benchmarks, with a detailed breakdown from its founder.
The tweet claims that the open-source Kimi K2.6 model has surpassed Claude Opus 4.7, marking a significant milestone for open-source AI in just three months. It provides a link to a full guide and prompts to verify the comparison.