Kimi-K3

45 records across 5 hardware.

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
MI355XKimi-K3 BF16vLLM1083.0–––tg10242026-09-27community–source
MI355XKimi-K3 BF16vLLM1082.6–––tg10242026-09-27community–source
GB300Kimi-K3 NVFP4vLLM946–––10242026-10-07sourced–source
GB300Kimi-K3 NVFP4vLLM942–––10242026-10-07sourced–source
GB300Kimi-K3 NVFP4vLLM909–––10242026-10-07sourced–source
GB300Kimi-K3 NVFP4vLLM902–––10242026-10-07sourced–source
8x MI355XKimi-K3 unspecifiedvLLM693.13––336.53–2026-10-05sourced–source
8x MI355XKimi-K3 unspecifiedvLLM685.62––340.08–2026-10-05sourced–source
MI355XKimi-K3 BF16vLLM668.97–––tg10242026-09-28community–source
MI355XKimi-K3 BF16vLLM658.42–––tg10242026-09-28community–source
MI355XKimi-K3 MXFP4vLLM648–––81922026-10-02community–source
MI355XKimi-K3 MXFP4vLLM647–––81922026-10-02community–source
MI355XKimi-K3 MXFP4vLLM646–––81922026-10-02community–source
MI355XKimi-K3 MXFP4vLLM623–––81922026-10-02community–source
MI355XKimi-K3 FP4vLLM529.7–––163842026-10-03community–source
MI355XKimi-K3 FP4vLLM515.1–––163842026-10-03community–source
MI355XKimi-K3 FP4vLLM510.9–––163842026-10-03community–source
MI355XKimi-K3 FP4vLLM499.3–––163842026-10-03community–source
MI355XKimi-K3 FP4vLLM485.5–––163842026-10-03community–source
MI355XKimi-K3 FP4vLLM453.4–––163842026-10-03community–source
GB300Kimi-K3 NVFP4vLLM415–––10242026-10-07sourced–source
GB300Kimi-K3 NVFP4vLLM413–––10242026-10-07sourced–source
MI355XKimi-K3 BF16vLLM236.05–––tg10242026-09-27community–source
MI355XKimi-K3 BF16vLLM236.05–––tg10242026-09-27community–source
MI355XKimi-K3 unstatedvLLM233.11–––10242026-10-13sourced–source
MI355XKimi-K3 unstatedvLLM232.87–––10242026-10-13sourced–source
MI355XKimi-K3 unstatedvLLM190.9––––2026-10-14sourced–source
MI355XKimi-K3 unstatedvLLM180.98––––2026-10-14sourced–source
MI355XKimi-K3 unstatedvLLM179.77–––10242026-10-13sourcedcontradictionsource
MI355XKimi-K3 unspecifiedvLLM177.3–––1000002026-10-07community–source
MI355XKimi-K3 unstatedvLLM175.7––––2026-10-14sourced–source
MI355XKimi-K3 unstatedvLLM174.25––––2026-10-14sourced–source
MI355XKimi-K3 unstatedvLLM172.89––––2026-10-06sourced–source
MI355XKimi-K3 unstatedvLLM164.79–––10242026-10-13sourcedcontradictionsource
MI355XKimi-K3 unstatedvLLM163.94––––2026-10-06sourced–source
8x MI355XKimi-K3 unstatedvLLM153.09––7291.12–2026-10sourced–source
8x MI355XKimi-K3 unstatedvLLM152.49––7280.8–2026-10sourced–source
8x MI355XKimi-K3 unstatedvLLM122.54––17312.48–2026-10sourced–source
8x MI355XKimi-K3 unstatedvLLM122.19––17272.42–2026-10sourced–source
GB300Kimi-K3 NVFP4vLLM78.7–––10242026-10-07sourced–source
GB300Kimi-K3 NVFP4vLLM78.2–––10242026-10-07sourced–source
GB300Kimi-K3 NVFP4vLLM29.6–––81922026-10-07sourced–source
GB300Kimi-K3 NVFP4vLLM29.4–––81922026-10-07sourced–source
8×H100 (640 GB HBM)Kimi-K3 IQ1_S-XSllama.cpp (CUDA)6.8–––81922026-08-10community–source
M3 UltraKimi-K3 mxfp4oMLX6.595.8––40962026-10-17community–source

← all models