Some users praise Kimi K3 for strong math performance and expected cloud speed gains, while many others criticize its reliability issues such as buffering, API queues, and lost interest after testing.
Based on 8 visible X reactions from 39 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@jasondeanlee Jason you need to get the Vivace plan. Kimi K3 is legit good at research level math.
@Yuchenj_UW I have good uptime via offical API (kimi for coding) but it can be very slow (10 Tok/s-25 Tok/s) but if you can deal with it the model is delivering great performance.
@Yuchenj_UW API availability has been the bottleneck for a while. If their inference fleet can't keep up with demand, the product experience suffers regardless of model quality.
@Yuchenj_UW benchmarks say one thing, the queue says another. the cheapest tokens on the leaderboard mean nothing if you cant get a response back
@jasondeanlee the fact that people were gassing up mira's garbo lab release. "inkling" LOL. just no
@Yuchenj_UW K3 still buffering after two days sounds rough.
I really want to try Kimi K3 on the Kimi web app, but I’ve tried repeatedly over the past 2 days and still haven’t gotten a response. How is everyone using K3? Is the API uptime good? Kimi needs a lot more GPUs for inference.
Inkling is crap and kimi is "task paused due to system peak" this is why we still pay oai and anthropic
Some users praise Kimi K3 for strong math performance and expected cloud speed gains, while many others criticize its reliability issues such as buffering, API queues, and lost interest after testing.
Based on 8 visible X reactions from 39 accounts; directional sample.
Ask a question below.
Published answers will appear here.