Flint — Hardware Requirements & GPU Compatibility
ChatFunctions
Flint is a mixture-of-experts language model for coding agents from Recursive. It is built to investigate codebases, fix bugs, implement features, and write tests, with reasoning and multi-turn tool use over text input and output. The card gives 39.8 billion total parameters, about 2.5 billion active per token, 256 routed experts with 6 selected per token, and a mix of sliding-window and full-attention layers. A Q4 build needs roughly 23 GB of memory, so it fits a 24 GB GPU only tightly and suits a 32 GB card or a unified-memory machine better. The documented context configuration is 262,144 tokens, shared between prompt and output. It is released under the Apache 2.0 license, which permits commercial use, and was published on October 21, 2026. Its card reports 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2, and 74.2% on SWE-bench Verified. The BF16 checkpoint is about 80 GB, and the card describes serving it with SGLang on one B300 or H200 GPU.
Usable on a 24 GB GPU
Runs from 16 GB (IQ2_M · 32K)Best on 24 GB (IQ4_NL · ~205 tok/s · 16K)Macs from 16 GB (IQ2_XXS · ~67 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Flint
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
ollama run hf.co/lmstudio-community/flint-GGUF:Q4_K_MQ4_K_M needs offload on the RTX 4090 · 1.0 GB short · 16K context
Download: flint-Q4_K_M.gguf24.17 GBlmstudio-community
Other ways
each opens step-by-stepRun in cloud
Fits on RTX PRO 4500 (32 GB) (7 GB headroom) · Q4_K_M
- Generation speed
- ~188 tok/s
- generation speed
- Cost per 1M output tokens
- $0.50
- per 1M output tokens
How Much VRAM Does Flint Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 17.3 GB | 18.9 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 128K context · ~230 tok/s | |
| 17.8 GB | 19.4 GB | 3-bit small — 3-bit small quantization | Fits: 64K context · ~227 tok/s | |
| 19.8 GB | 21.3 GB | 3-bit medium — 3-bit medium quantization | Fits: 64K context · ~218 tok/s | |
| 20.3 GB | 21.8 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 64K context · ~215 tok/s | |
| 24.3 GB | 25.8 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Does not fit: needs 25 GB | |
| 28.8 GB | 30.3 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Does not fit: needs 29.5 GB | |
| 33.2 GB | 34.8 GB | 6-bit, very good — 6-bit quantization, very good quality | Does not fit: needs 34 GB | |
| 40.2 GB | 41.7 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 40.9 GB |
Selected Q4_K_M
4-bit medium quantization25.0GBat 16K context
- Weights
- 23.9 GB
- KV cache
- 0.8 GB
- Overhead
- 0.3 GB
Doesn't fit RTX 4090 (reference) · needs 1.0 GB more, IQ4_NL fits
≥50 tok/s needs ≥135 GB/s · RTX 5090 ~250, NVIDIA DGX Spark ~89
Which Hardware Can Run Flint?
Q4_K_M · 24.3 GBFlint (Q4_K_M) requires 24.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 32+ GB is recommended. Using the full 262K context window can add up to 13.1 GB, bringing total usage to 37.3 GB. 1 GPU can run it, including NVIDIA GeForce RTX 5090. 98 devices with unified memory can run Flint, including NVIDIA DGX Spark, NVIDIA Jetson AGX Thor Developer Kit, NVIDIA Jetson AGX Orin 32GB.
20 GB
Q4_K_M · best fit IQ3_M (19.0 GB)
1 GPU
RX 7900 XT~82Simulate output on RX 7900 XT32–40 GB
fits · 7.0 GB free · up to 128K ctx
1 GPU · 3 Mac classes · 2 device families
RTX 5090~250Simulate output on RTX 5090M5 Max36 GB~69Simulate output on Mac Studio M5 Max (36 GB)M532 GB~42Simulate output on MacBook Air 13" M5 (32 GB)Orin~71Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~45Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Similar models
2| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Phi 3.5 MoE InstructMicrosoft | — | 23.9 GB | ~141 | same size |
| Falcon 40B InstructTII UAE | — | 23.8 GB | ~33 | same size |
Specifications
- Publisher
- RecursiveSI
- Parameters
- 39.8B
- Active parameters
- 2.5B per token
- Architecture
- FlintForCausalLM
- Context Length
- 262,144 tokens
- Vocabulary Size
- 150,144
- Release Date
- 2026-10-21
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Flint need?
Flint requires 24.3 GB of VRAM at Q4_K_M, or 80.0 GB at BF16. Full 262K context adds up to 13.1 GB (37.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 39.8B × 4.8 bits ÷ 8 = 23.9 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 ≈ 13.4 GB (at full 262K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run Flint?
Yes, at Q4_K_S (22.8 GB) or lower. Higher quantizations like Q4_K_M (24.3 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for Flint?
For Flint, Q4_K_M (24.3 GB) offers the best balance of quality and VRAM usage. Q4_K_L (24.8 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 11.3 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Flint on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Flint, the smallest at IQ2_XXS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 24.3 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio M1 Max (32 GB).
Can I run Flint locally?
Yes — Flint can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Flint?
At Q4_K_M, Flint can reach ~99 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 1.5 GB active ÷ (8000 × 0.65) = 0.293 ms, plus 49 layers × 0.055 ms = 2.695 ms, so 1000 ÷ 2.988 ms = ~335 tok/s
Estimated speed at Q4_K_M (24.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Flint?
At Q4_K_M, the download is about 23.87 GB. The full-precision BF16 version is 79.57 GB. The smallest option (IQ2_XXS) is 10.94 GB.
Which GPUs can run Flint?
1 consumer GPU can run Flint at Q4_K_M. Top options include NVIDIA GeForce RTX 5090.
Which devices can run Flint?
98 devices with unified memory can run Flint at Q4_K_M (24.3 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (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.