Runtime recipe
8 × H200 SXM
141 GB per GPU
INT4 experts / mixed
vLLM model-specific guide
TP8 on an H200 node; multimodal encoder placement and request state still consume memory. Follow the exact checkpoint loader.
Kimi / MoE
Kimi-K2.6
Keep the hardware fixed; compare the completed work
New to model sizes or GPU memory? Start with weights, parameters and quantization →
K2.6 is a useful controlled upgrade candidate for an existing K2.5 service. Its total and active counts remain 1T and 32B, so the adoption decision should turn on your tasks, tool reliability and response lengths.
1,000Breported total parameters
32Bactive parameters per token
262,144configured token positions
Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size.
Evaluate it for
Existing K2.5 deployments evaluating agent and multimodal task improvements.
Choose another path when
A new small-GPU service for which the trillion-parameter store is the main obstacle.
Release context
Why this release matters
This is a later K2 multimodal generation, not K3’s hybrid-attention redesign. Matching architecture dimensions makes a fair deployment comparison easier; it does not establish identical weights or quality.
Deployment starting points
Which hardware, how many, at what precision?
Start with an exact artifact and an engine that supports it. A configuration below is evidence of a documented path; its device count does not promise a particular throughput or concurrency.
Runtime recipe
8 × MI355X
288 GB per GPU
W4A16 experts / mixed
vLLM + ROCm 7.2.1
Use AITER and the guide’s gfx950/FlyDSL settings. This is a separate AMD kernel path.
595.15 GB of tensor data in the pinned index
This is the index’s reported tensor payload, in decimal GB. It excludes file headers and runtime memory. Mixed precision, conversion and host/device placement determine how much GPU memory the loaded model needs.
Inspect the exact index metadata →Capacity screen: would the weights fit?
This estimates inference memory for a hypothetical uniform precision. It is useful for rejecting an allocation that is too small. It does not establish a working deployment or estimate training memory.
A uniform-precision GPU count would hide this model’s component layout. Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size. Use the documented artifact and placement path above; no GPU count is inferred here.
Architecture in practice
What the serving engine has to do
Most experts are stored. Only a few execute.
Compute follows selected experts.
Weight memory follows the full expert store.
Mechanism schematic based on the pinned model card and configuration. It explains a design principle; it is not a full implementation graph.
Inspect the full checkpoint specifications
| Exact checkpoint | moonshotai/Kimi-K2.6 |
|---|---|
| Text model type | kimi_k2 |
| Layers | 61 |
| Hidden width | 7,168 |
| Attention / KV heads | 64 / 64 |
| Head dimension | Not verified |
| Routed / selected experts | 384 / 8 |
| Configured positions | 262,144 |
| Documented extension | Not recorded for this checkpoint |
| Inputs → output | text, image, video → text |
| License metadata | modified-mit · custom terms |
Open weights do not imply unrestricted use. Read the applicable license terms linked from the pinned card.
A useful comparison
What else belongs on the shortlist?
K2.5 is the upgrade baseline; K2.7-Code is the more focused coding comparison.
27 Jan 2026
Kimi-K2.5
The multimodal starting point for the K2 series
1,000B reported total · text + image + video
12 Jun 2026
Kimi-K2.7-Code
A coding upgrade without a new size class
1,000B reported total · text + image + video
These are editorial comparison candidates. We have not run a matched quality or serving benchmark, so this is not a ranking.
Prepare to explain it
The interview lesson
Hold context, concurrency and harness fixed, then compare successful tasks and tokens consumed per success.
Attention layouts · Quantization · Courses and worked examples
For a deployment evaluation, record exact weights, precision, engine, device count, interconnect, prompt/output lengths and concurrency. Report task success, errors, TTFT, TPOT and useful throughput together.
Sources and citation
Our assessment is editorial judgment based on the linked architecture and deployment evidence. Model facts are publisher-reported or attributed to runtime maintainers; no independent GPU benchmark was run.
AI Infra Interviews, “Kimi-K2.6: release, architecture and deployment”, checked 2026-09-13. Preserve this date and the exact checkpoint when citing.