Recursive/Mixture of Experts

RFlint

Name: Flint
Author: Recursive

Flint — Recursive's first model, a 40B MoE for coding agents with about 2.5B active, trained from scratch. Most of its attention layers use a 513-token sliding window, which keeps the KV cache small; the full-precision checkpoint is about 80GB, so consumer hardware needs one of the 4-bit builds.

chatcodingreasoningagentictool_use

39.79B

Parameters (2.5B active)

256K

Context length

Benchmarks

Quantizations

33K

HF downloads

Architecture

MoE

Released

2026-10-21

Layers

KV Heads

Head Dim

128

Family

other

Quantization Options

Context length:

Quant	Bits	VRAM @ 16K	Quality
IQ3_XXS	3.25	17.1 GB 16.7 + 0.4 KV	low
IQ3_XS	3.5	18.3 GB 17.9 + 0.4 KV	low
Q3_K_S	3.64	19.0 GB 18.6 + 0.4 KV	low
IQ3_M	3.76	19.6 GB 19.2 + 0.4 KV	low
Q3_K_M	4	20.8 GB 20.4 + 0.4 KV	low
Q3_K_L	4.3	22.3 GB 21.9 + 0.4 KV	moderate
IQ4_XS	4.46	23.1 GB 22.7 + 0.4 KV	moderate
Q4_K_S	4.67	24.1 GB 23.7 + 0.4 KV	moderate
Q4_K_M	4.89	25.2 GB 24.8 + 0.4 KV	good
Q5_K_S	5.57	28.6 GB 28.2 + 0.4 KV	good
Q5_K_M	5.7	29.2 GB 28.8 + 0.4 KV	good
Q6_K	6.56	33.5 GB 33.1 + 0.4 KV	excellent
Q8_0	8.5	43.2 GB 42.8 + 0.4 KV	lossless
FP16	16	80.5 GB 80.1 + 0.4 KV	lossless

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

Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise →

▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN FLINT 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 (2)

LiveCodeBench80.3

SWE-bench74.2

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

curl -fsSL https://ollama.com/install.sh | sh

$ollama run flint:40b

Tag may need adjustment — check ollama.com/library/flint 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

HuggingFace Ollama Library GGUF Downloads Build Hardware

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

AMD Radeon PRO V710

28 GB VRAM • 504 GB/s

AMD

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA RTX 5090

32 GB VRAM • 1792 GB/s

NVIDIA

$1999

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

Apple M1 Max (32GB)

32 GB VRAM • 400 GB/s

APPLE

$1499

Apple M2 Max (32GB)

32 GB VRAM • 400 GB/s

APPLE

$1799

NVIDIA V100 SXM2 32GB

32 GB VRAM • 900 GB/s

NVIDIA

$3500

Apple M2 Pro (32GB)

32 GB VRAM • 200 GB/s

APPLE

$1499

Apple M4 (32GB)

32 GB VRAM • 120 GB/s

APPLE

$1199

NVIDIA Tesla V100 DGXS 32 GB

32 GB VRAM • 897 GB/s

NVIDIA

NVIDIA Tesla V100 PCIe 32 GB

32 GB VRAM • 897 GB/s

NVIDIA

NVIDIA Tesla V100 SXM2 32 GB

32 GB VRAM • 898 GB/s

NVIDIA

NVIDIA Tesla V100 SXM3 32 GB

32 GB VRAM • 981 GB/s

NVIDIA

AMD Radeon Instinct MI60

32 GB VRAM • 1020 GB/s

AMD

NVIDIA Tesla V100S PCIe 32 GB

32 GB VRAM • 1130 GB/s

NVIDIA

AMD Radeon Instinct MI100

32 GB VRAM • 1230 GB/s

AMD

$5000

NVIDIA RTX 5000 Ada Generation

32 GB VRAM • 576 GB/s

NVIDIA

$4000

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA GeForce RTX 5090

32 GB VRAM • 1790 GB/s

NVIDIA

$1999

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA GeForce RTX 5090 D

32 GB VRAM • 1790 GB/s

NVIDIA

$1999

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA Jetson AGX Xavier 32 GB

32 GB VRAM • 136 GB/s

NVIDIA

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA Quadro GV100

32 GB VRAM • 868 GB/s

NVIDIA

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA TITAN V CEO Edition

32 GB VRAM • 868 GB/s

NVIDIA

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA Tesla PG500-216

32 GB VRAM • 1130 GB/s

NVIDIA

NVIDIA Tesla PG503-216

32 GB VRAM • 1130 GB/s

NVIDIA

AMD Radeon Pro Vega II

32 GB VRAM • 825 GB/s

AMD

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

AMD Radeon Pro Vega II Duo

32 GB VRAM • 1020 GB/s

AMD

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

AMD Radeon PRO V620

32 GB VRAM • 512 GB/s

AMD

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

AMD Radeon PRO W6800

32 GB VRAM • 512 GB/s

AMD

$2249

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

AMD Radeon Pro W6800X

32 GB VRAM • 512 GB/s

AMD

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

AMD Radeon Pro W6800X Duo

32 GB VRAM • 512 GB/s

AMD

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

AMD Radeon Pro W6900X

32 GB VRAM • 512 GB/s

AMD

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

NVIDIA Jetson AGX Orin 32 GB

32 GB VRAM • 205 GB/s

NVIDIA

Amazon Newegg eBay PCPartPicker Rent on RunPod Rent on Vast.ai

Find the best GPU for Flint

Build Hardware for Flint

QUANT	BPW	VRAM	QUALITY
IQ3_XXS	3.25	16.7 GB	82%
IQ3_XS	3.5	17.9 GB	84%
Q3_K_S	3.64	18.6 GB	85%
IQ3_M	3.76	19.2 GB	86%
Q3_K_M	4	20.4 GB	88%
Q3_K_L	4.3	21.9 GB	90%
IQ4_XS	4.46	22.7 GB	92%
Q4_K_S	4.67	23.7 GB	93%
Q4_K_M	4.89	24.8 GB	94%
Q5_K_S	5.57	28.2 GB	96%
Q5_K_M	5.7	28.8 GB	96%
Q6_K	6.56	33.1 GB	97%
Q8_0	8.5	42.8 GB	100%
FP16	16	80.1 GB	100%

RFlint

Quantization Options

Community Ratings

Benchmarks (2)

Run this model

Auto-setup with fitmyllm CLI

GPUs that can run this model

Flint — 39.79B MoE.