Miau Labs / Insight
Kimi K3, and what we can still learn from the pelican benchmark
Moonshot AI has announced Kimi K3, a 2.8 trillion parameter AI model that surpasses Claude Fable 5 and GPT-5.6 Sol in some benchmarks. The model is available via the Moonshot AI website and API, with an open weight release promised by July 27, 2026. The pricing of the model is notable, with a cost of $3 per million inp
Moonshot AI has announced Kimi K3, a 2.8 trillion parameter AI model that surpasses Claude Fable 5 and GPT-5.6 Sol in some benchmarks. The model is available via the Moonshot AI website and API, with an open weight release promised by July 27, 2026. The pricing of the model is notable, with a cost of $3 per million input tokens and $15 per million output tokens, making it the most expensive model released by a Chinese AI lab. The pelican benchmark is no longer a reliable indicator of a model's capabilities, and agentic tool calling and the ability to operate tools reliably are more important for today's models.
Moonshot AI has announced Kimi K3, a 2.8 trillion parameter AI model that surpasses Claude Fable 5 and GPT-5.6 Sol in some benchmarks. The model is available via the Moonshot AI website and API, with an open weight release promised by July 27, 2026. The pricing of the model is notable, with a cost of $3 per million inp
- Moonshot AI has announced Kimi K3, a 2.8 trillion parameter AI model that surpasses Claude Fable 5 and GPT-5.6 Sol in some benchmarks.
- Kimi K3 is available via the Moonshot AI website and API, with an open weight release promised by July 27, 2026.
- The model is notable for its pricing, with a cost of $3 per million input tokens and $15 per million output tokens, making it the most expensive model released by a Chinese AI lab.
Miau Labs takeThe pelican benchmark is no longer a reliable indicator of a model's capabilities, and agentic tool calling and the ability to operate tools reliably are more important for today's models.