Kimi K2.6 Eagle3 Mla — Hardware Requirements & GPU Compatibility
Chat
Kimi K2.6 Eagle3 Mla is a 3.0B-parameter open language model from lightseekorg in the Kimi K2 family. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 2.17 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 128K)Best on 24 GB (Q4_K_M · ~302 tok/s · 256K)Macs from 8 GB (Q4_K_M · ~33 tok/s)
Other
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Kimi K2.6 Eagle3 Mla
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
vllm serve lightseekorg/kimi-k2.6-eagle3-mla --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX 3080 Ti (12 GB) (9 GB headroom) · Q4_K_M
- Generation speed
- ~273 tok/s
- generation speed
- Cost per 1M output tokens
- $0.12
- per 1M output tokens
How Much VRAM Does Kimi K2.6 Eagle3 Mla Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 1.6 GB | 2.5 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 256K context · ~400 tok/s | |
| 1.8 GB | 2.7 GB | 3-bit medium — 3-bit medium quantization | Fits: 256K context · ~358 tok/s | |
| 2.2 GB | 3.1 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 256K context · ~302 tok/s | |
| 2.5 GB | 3.4 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 256K context · ~261 tok/s | |
| 2.9 GB | 3.7 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 256K context · ~230 tok/s | |
| 3.4 GB | 4.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 256K context · ~194 tok/s | |
| 6.4 GB | 7.3 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 256K context · ~103 tok/s |
Selected Q4_K_M
4-bit medium quantization2.6GBat 16K context
- Weights
- 1.8 GB
- KV cache
- 0.5 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 21.4 GB free · ~302 tok/s
≥50 tok/s needs ≥167 GB/s · RTX 5090 ~537, RTX 3090 Ti ~302
Which Hardware Can Run Kimi K2.6 Eagle3 Mla?
Q4_K_M · 2.2 GBKimi K2.6 Eagle3 Mla (Q4_K_M) requires 2.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 3+ GB is recommended. Using the full 262K context window can add up to 7.5 GB, bringing total usage to 9.6 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run Kimi K2.6 Eagle3 Mla, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB).
≤ 11 GB
≤ 11 GB tier →fits · 5.4 GB free · up to 128K ctx
16 GPUs · 1 Mac class · 2 device families
RTX 3080~228Simulate output on RTX 3080RTX 2080 Ti~185Simulate output on RTX 2080 TiRTX 3070 Ti~182Simulate output on RTX 3070 TiM3*8 GB~33Simulate output on iMac 24" M3 (8 GB)Orin Nano~31Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)iPhone~22Simulate output on iPhone 1720 GB
fits · 17.4 GB free · up to 256K ctx
1 GPU
RX 7900 XT~221Simulate output on RX 7900 XT32–40 GB
fits · 29.4 GB free · up to 256K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~537Simulate output on RTX 5090M5 Max36 GB~148Simulate output on Mac Studio M5 Max (36 GB)M532 GB~50Simulate output on MacBook Air 13" M5 (32 GB)Orin~61Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~40Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to Kimi K2.6 Eagle3 Mla
Kimi K2 family · 8 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| D1 3BLiquid AI | — | 2.6 GB | ~285 | same size |
| SmolLM3 3BHugging Face | — | 3.4 GB | ~285 | same size |
| Llama 3.2 3B InstructMeta | — | 4.1 GB | ~309 | same size |
| AI21 Jamba Reasoning 3BAI21 Labs | — | 2.3 GB | ~291 | same size |
| Qwen2.5 3B InstructAlibaba | — | 2.8 GB | ~294 | same size |
| Quasar 3B A1B Previewsilx-ai | — | 8.4 GB | ~167 | same size |
Specifications
- Publisher
- lightseekorg
- Family
- Kimi K2
- Parameters
- 3.0B
- Architecture
- Eagle3DeepseekV2ForCausalLM
- Context Length
- 262,144 tokens
- Vocabulary Size
- 163,840
- Release Date
- 2026-04-30
- License
- Other
Frequently Asked Questions
How much VRAM does Kimi K2.6 Eagle3 Mla need?
Kimi K2.6 Eagle3 Mla requires 2.2 GB of VRAM at Q4_K_M, or 6.4 GB at BF16. Full 262K context adds up to 7.5 GB (9.6 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 3.0B × 4.8 bits ÷ 8 = 1.8 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 7.8 GB (at full 262K context)
VRAM usage by quantization
What's the best quantization for Kimi K2.6 Eagle3 Mla?
For Kimi K2.6 Eagle3 Mla, Q4_K_M (2.2 GB) offers the best balance of quality and VRAM usage. Q5_K_M (2.5 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 1.6 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Kimi K2.6 Eagle3 Mla on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 145 other Macs can run Kimi K2.6 Eagle3 Mla. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 2.2 GB of usable unified memory (RAM minus macOS overhead).
Can I run Kimi K2.6 Eagle3 Mla locally?
Yes — Kimi K2.6 Eagle3 Mla can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Kimi K2.6 Eagle3 Mla?
At Q4_K_M, Kimi K2.6 Eagle3 Mla can reach ~2212 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~302 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 2.17 × 0.65 = ~2396 tok/s
Estimated speed at Q4_K_M (2.2 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Kimi K2.6 Eagle3 Mla?
At Q4_K_M, the download is about 1.81 GB. The full-precision BF16 version is 6.03 GB. The smallest option (Q2_K) is 1.28 GB.
Which GPUs can run Kimi K2.6 Eagle3 Mla?
52 consumer GPUs can run Kimi K2.6 Eagle3 Mla at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run Kimi K2.6 Eagle3 Mla?
163 devices with unified memory can run Kimi K2.6 Eagle3 Mla at Q4_K_M (2.2 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.