Updated minutes ago
ReleasedJune 11, 2026Verified — · huggingface.co
Moonshot

Kimi K2.7 Code

Moonshot AI · moe · 1026.9B parameters · 262,144 context

Quality
50.0
Deploy this model →Use via API →Calculate ROI →Compare with othersFine-tune this model

Parameters

1.0T

Context Window

256K tokens

Architecture

MoE

Best GPU

B200 NVL (pair)

Cheapest API

$3.50/M

Intelligence Brief

Kimi K2.7 Code is a 1026.9B parameter Mixture-of-Experts (384 experts, 8 active) model from Moonshot AI, featuring Multi-Head Attention (MHA) with 61 layers and 7,168 hidden dimensions. With a 262,144 token context window, it supports vision, code. The most cost-effective API deployment is via openrouter at $3.50/M output tokens. For self-hosted inference, B200 NVL (pair) delivers optimal throughput at $39858/month.

Provider pricing

1 provider · canonical: openrouter
Provider Input $/M Output $/M ▲Notes
openroutercanonical$0.700$3.50cheapest input · cheapest output

Prices update via the nightly pricing cron + admin approvals at /admin/ingest-queue. The leaderboard's Input/Output cells show the canonical rate above; this table shows the full spread.

Recent changes

Loading…

Related models

5 suggestions

Picks: same family first, then same vendor within ±2× params, then top tag-overlap matches. Price shown is the cheapest Output $/M across providers — the row's page shows the canonical anchor.

Architecture Details

TypeMOE
Total Parameters1026.9B
Active Parameters32.9B
Layers61
Hidden Dimension7,168
Attention Heads64
KV Heads64
Head Dimension112
Vocab Size163,840
Total Experts384
Active Experts8

Memory Requirements

BF16 Weights

1912.7 GB

FP8 Weights

956.4 GB

INT4 Weights

478.2 GB

KV-Cache per Token1748992 bytes
Activation Estimate0.00 GB

Fits on (multi-GPU with Tensor Parallelism)

RECOMMENDEDx2B200 NVL (pair)360GB eachINT4x3Groq LPU230GB eachINT4x3Instinct MI325X256GB eachINT4x3B300288GB eachINT4x4B200 SXM180GB eachINT4x4H100 NVL 94GB (per GPU pair)188GB eachINT4x4B100 SXM192GB eachINT4x4GB200 NVL72 (per GPU)192GB eachINT4x4GB300 NVL72 (per GPU)192GB eachINT4x4Instinct MI300X192GB eachINT4x5Instinct MI250X128GB eachINT4x5Gaudi 3128GB eachINT4

Multi-GPU configurations use Tensor Parallelism (TP) to split model layers across GPUs. Requires NVLink or NVSwitch interconnect for optimal performance.

This model requires multi-GPU deployment. Minimum: 2x B200 NVL (pair) (360GB each) with Tensor Parallelism.

GPU Compatibility Matrix

Kimi K2.7 Code is compatible with 0% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
✗
✗
✗
B300288GB
✗
✗
✗
B100 SXM192GB
✗
✗
✗
GB200 NVL72 (per GPU)192GB
✗
✗
✗
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
✗
✗
✗
H200 SXM141GB
✗
✗
✗
H2096GB
✗
✗
✗
GH20096GB
✗
✗
✗
Ada Lovelace(11 GPUs)
L40S48GB
✗
✗
✗
L4048GB
✗
✗
✗
RTX 6000 Ada48GB
✗
✗
✗
L2048GB
✗
✗
✗
Ampere(16 GPUs)
A100 80GB SXM80GB
✗
✗
✗
A100 80GB PCIe80GB
✗
✗
✗
A1664GB
✗
✗
✗
RTX A600048GB
✗
✗
✗
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

B200 NVL (pair)optimal

FP8 · 4 GPUs · tensorrt-llm

98/100

score

Throughput

140.0 tok/s

Latency (ITL)

7.1ms

Est. TTFT

1ms

Cost/Month

$39858

Cost/M Tokens

$108.33

Use this config →
B200 SXMoptimal

FP8 · 8 GPUs · tensorrt-llm

93/100

score

Throughput

140.0 tok/s

Latency (ITL)

7.1ms

Est. TTFT

1ms

Cost/Month

$34088

Cost/M Tokens

$92.65

Use this config →
B100 SXMoptimal

FP8 · 8 GPUs · tensorrt-llm

93/100

score

Throughput

140.0 tok/s

Latency (ITL)

7.1ms

Est. TTFT

1ms

Cost/Month

$34164

Cost/M Tokens

$92.86

Use this config →

Deployment Options

API

API Deployment

openrouter

$3.50/M

output tokens

Self-Hosted

Single GPU

Requires multi-GPU setup (956 GB VRAM needed)

Scale

Multi-GPU

B200 NVL (pair) x4

140.0 tok/s

TP· $39858/mo

Open Full Calculator →

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
openrouter$0.70$3.50
Cheapest

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
openrouterBest Value$0.70$3.50$21

Cost per 1,000 Requests

Short (500 tok)

$1.05

via openrouter

Medium (2K tok)

$4.20

via openrouter

Long (8K tok)

$12.60

via openrouter

Performance Estimates

Throughput by GPU

B200 NVL (pair)
140.0 tok/s
B200 SXM
140.0 tok/s
B100 SXM
140.0 tok/s

VRAM Breakdown (B200 NVL (pair), FP8)

Weights
Weights 256.7 GBKV-Cache 14.3 GBActivations 0.0 GBOverhead 12.8 GB

Precision Impact

bf16

513.5 GB

weights/GPU

fp8

256.7 GB

weights/GPU

~140.0 tok/s

int4

128.4 GB

weights/GPU

Capabilities

Features

✗ Tool Use✓ Vision✓ Code✗ Math✗ Reasoning✗ Multilingual✗ Structured Output

Supported Frameworks

vllmsglang

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy Kimi K2.7 Code

Similar Models

Kimi K2.5

1000B params · moe

Quality: 54

from $0.00/M

CheaperCompare →

Kimi K3

2779.9B params · moe

Quality: 50

from $15.00/M

Larger context, More expensive, Larger modelCompare →

GPT-5.5

700B params · moe

Quality: 50

from $30.00/M

Larger context, More expensiveCompare →

DeepSeek V3-0324

685B params · moe

Quality: 81

from $0.00/M

Higher quality, CheaperCompare →

DeepSeek R1

671B params · moe

Quality: 88

from $2.00/M

Higher quality, CheaperCompare →

Frequently Asked Questions

How much VRAM does Kimi K2.7 Code need for inference?

Kimi K2.7 Code requires approximately 1912.7 GB of VRAM at BF16 precision, 956.4 GB at FP8, or 478.2 GB at INT4 quantization. Additional VRAM is needed for KV-cache (1748992 bytes per token) and activations (~0.00 GB).

What is the best GPU for Kimi K2.7 Code?

The top recommended GPU for Kimi K2.7 Code is the B200 NVL (pair) (x4) using FP8 precision. It achieves approximately 140.0 tokens/sec at an estimated cost of $39858/month ($108.33/M tokens). Score: 98/100.

How much does Kimi K2.7 Code inference cost?

Kimi K2.7 Code API inference starts from $0.70/M input tokens and $3.50/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.