FitMyLLM
Ant Group/Mixture of Experts

ALing 3.1 Flash 1019

Ling 3.1 Flash 1019 is Ant Group's October re-post-train of Ling 3.1 Flash: a 560B MoE with 25B active, hybrid Kimi Delta Attention and gated MLA at 7:1, so only 9 of 72 layers keep a KV cache. It thinks by default and is built for agentic and coding work.

chatcodingreasoningmultilingualmathagentictool_use
572.12B
Parameters (25B active)
256K
Context length
4
Benchmarks
17
Quantizations
2K
HF downloads
Architecture
MoE
Released
2026-10-27
Layers
72
KV Heads
64
Head Dim
128
Family
other

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
IQ2_XXS2.38
170.8 GB
170.7 + 0.1 KV
low
IQ2_M2.93
210.1 GB
210.0 + 0.1 KV
low
Q2_K3.16
226.6 GB
226.5 + 0.1 KV
low
IQ3_XXS3.25
233.0 GB
232.9 + 0.1 KV
low
IQ3_XS3.5
250.9 GB
250.8 + 0.1 KV
low
Q3_K_S3.64
260.9 GB
260.8 + 0.1 KV
low
IQ3_M3.76
269.5 GB
269.4 + 0.1 KV
low
Q3_K_M4
286.7 GB
286.5 + 0.1 KV
low
Q3_K_L4.3
308.1 GB
308.0 + 0.1 KV
moderate
IQ4_XS4.46
319.6 GB
319.4 + 0.1 KV
moderate
Q4_K_S4.67
334.6 GB
334.5 + 0.1 KV
moderate
Q4_K_M4.89
350.3 GB
350.2 + 0.1 KV
good
Q5_K_S5.57
398.9 GB
398.8 + 0.1 KV
good
Q5_K_M5.7
408.2 GB
408.1 + 0.1 KV
good
Q6_K6.56
469.7 GB
469.6 + 0.1 KV
excellent
Q8_08.5
608.5 GB
608.4 + 0.1 KV
lossless
FP1616
1144.8 GB
1144.7 + 0.1 KV
lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Too big for a single GPU — plan a multi-GPU deployment
Even the lightest quant needs ~171 GB. Size GPUs, replicas, TCO and scaling for a production setup. Open in Enterprise →
▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN LING 3.1 FLASH 1019 NOW

Rent on Vast.ai →Or RunPod →

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (4)

AA Long Context82.3
SciCode56.8
HLE43.5
Terminal-Bench 432.8

Run this model

▸Easiest way to get started·Beginners
DOCS ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run other:572b-q4_K_M

Tag may need adjustment — check ollama.com/library/other for available tags.

▸ SETUP GUIDE
>_

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more
Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines
HuggingFaceGGUF DownloadsBuild Hardware

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

Apple M4 Ultra (384GB)
384 GB VRAM • 1092 GB/s
APPLE
$9999
Apple M5 Ultra (384GB)
384 GB VRAM • 1228 GB/s
APPLE

Find the best GPU for Ling 3.1 Flash 1019

Build Hardware for Ling 3.1 Flash 1019
▸ SPEC SHEET

Ling 3.1 Flash 1019 — 572.12B MoE.

▸ SPECIFICATIONS
PARAMETERS
572.12B (25B active)
ARCHITECTURE
Mixture of Experts
CONTEXT LENGTH
256K tokens
CAPABILITIES
chat, coding, reasoning, multilingual, math, agentic, tool_use
RELEASE DATE
2026-10-27
PROVIDER
Ant Group
FAMILY
other
▸ VRAM REQUIREMENTS
QUANTBPWVRAMQUALITY
IQ2_XXS2.38170.7 GB65%
IQ2_M2.93210.0 GB75%
Q2_K3.16226.5 GB78%
IQ3_XXS3.25232.9 GB82%
IQ3_XS3.5250.8 GB84%
Q3_K_S3.64260.8 GB85%
IQ3_M3.76269.4 GB86%
Q3_K_M4286.5 GB88%
Q3_K_L4.3308.0 GB90%
IQ4_XS4.46319.4 GB92%
Q4_K_S4.67334.5 GB93%
Q4_K_M4.89350.2 GB94%
Q5_K_S5.57398.8 GB96%
Q5_K_M5.7408.1 GB96%
Q6_K6.56469.6 GB97%
Q8_08.5608.4 GB100%
FP16161144.7 GB100%
§ 01BENCHMARK SCORES
HLE43.5
aa_scicode56.8
aa_lcr82.3
aa_terminal_bench_v432.8
§ 03COMPATIBLE GPUs
2 @ Q4_K_M
Apple M4 Ultra (384GB)
384 GB · 1092 GB/s
Apple M5 Ultra (384GB)
384 GB · 1228 GB/s