It feels like releasing models right now only goes to two extremes:
Either you make the cheapest model as powerful as possible, or you go straight for the cutting-edge model—otherwise there’s simply no discussion and no real usage.
The two releases last night: Grok 4.7 and Mimo V2.6. Their capabilities are pretty much the same across various third-party benchmarks, but the price difference between them is way too big.
With Mimo V2.6 Pro, when there’s a cache hit, the input price is ¥0.025 and the output is ¥6. Meanwhile, if you convert Grok 4.7 from USD, the input is already ¥3.35, and the output is as high as ¥40.
After reading some of the feedback and reviews on Twitter this morning about Grok 4.7, the main complaints are that it’s expensive, and another that it might not perform that well—being slow and not very good.
Mimo’s price is more than twice cheaper than DeepSeek V4 (whether Flash or Pro), even at the current low point.
For example, Mimo V2.6 Flash’s output price is ¥2, while DeepSeek V4.1 Flash’s output price is ¥4.
Also, it seems Mimo V2.6 is scheduled to be free for a limited time on OpenCode starting next week.
Mimo has also launched its own desktop app. It seems to have a separate subscription price—you can give it a try. The desktop app includes an image generation model, so you can use it to sketch or design some interfaces, for example.
My guess is that both companies think OpenAI might release a new model this afternoon (maybe GPT-6 Sol and GPT-6 Luna, which apparently are very cheap). They wanted to launch before OpenAI, and the timing ended up overlapping.
Either you make the cheapest model as powerful as possible, or you go straight for the cutting-edge model—otherwise there’s simply no discussion and no real usage.
The two releases last night: Grok 4.7 and Mimo V2.6. Their capabilities are pretty much the same across various third-party benchmarks, but the price difference between them is way too big.
With Mimo V2.6 Pro, when there’s a cache hit, the input price is ¥0.025 and the output is ¥6. Meanwhile, if you convert Grok 4.7 from USD, the input is already ¥3.35, and the output is as high as ¥40.
After reading some of the feedback and reviews on Twitter this morning about Grok 4.7, the main complaints are that it’s expensive, and another that it might not perform that well—being slow and not very good.
Mimo’s price is more than twice cheaper than DeepSeek V4 (whether Flash or Pro), even at the current low point.
For example, Mimo V2.6 Flash’s output price is ¥2, while DeepSeek V4.1 Flash’s output price is ¥4.
Also, it seems Mimo V2.6 is scheduled to be free for a limited time on OpenCode starting next week.
Mimo has also launched its own desktop app. It seems to have a separate subscription price—you can give it a try. The desktop app includes an image generation model, so you can use it to sketch or design some interfaces, for example.
My guess is that both companies think OpenAI might release a new model this afternoon (maybe GPT-6 Sol and GPT-6 Luna, which apparently are very cheap). They wanted to launch before OpenAI, and the timing ended up overlapping.
