Ring Flash 2.0 — Hardware Requirements & GPU Compatibility
Chat
Ring-flash-2.0 is Inclusion AI's reasoning ("thinking") model, a mixture-of-experts system built on the same Ling 2.0 architecture and Ling-flash-base-2.0 checkpoint as Ling-flash-2.0, sharing its roughly 102.9-billion total and about 6.15-billion active parameter footprint. It adds long chain-of-thought supervised fine-tuning followed by reinforcement learning with verifiable rewards and a further RLHF stage, using Inclusion AI's own IcePop algorithm to stabilize MoE reinforcement learning against training-inference probability mismatches during long rollouts. The company reports it leads comparable open thinking models on math, code, and logical-reasoning benchmarks while retaining the creative-writing strength of its non-thinking twin, Ling-flash-2.0. Its small active-parameter footprint keeps generation fast, but all of its parameters must stay in memory, so it needs a multi-GPU setup or a high-memory workstation even once quantized. Context length is natively 32,768 tokens, extendable to 128,000 tokens with YaRN. It is released under the MIT license, permitting unrestricted commercial and research use. It was published in September 2025, alongside its non-reasoning sibling Ling-flash-2.0.
Doesn't fit a 24 GB GPU
Smallest consumer fit → RTX 5090 · IQ2_XXS · 29.7 GB · ~309 tok/s
Macs from 36 GB (IQ2_XS · ~82 tok/s)
MIT
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Ring Flash 2.0
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
ollama run hf.co/mradermacher/Ring-flash-2.0-i1-GGUFQ4_K_M · needs ~63.1 GB · 16K context
Download: Ring-flash-2.0-i1-Q4_K_M.gguf61.73 GBmradermacher
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on A100 80GB (16 GB headroom) · Q4_K_M
- Generation speed
- ~219 tok/s
- generation speed
- Cost per 1M output tokens
- $1.44
- per 1M output tokens
How Much VRAM Does Ring Flash 2.0 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 28.7 GB | 30.7 GB | 2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss | Does not fit: needs 29.7 GB | |
| 44.2 GB | 46.2 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Does not fit: needs 45.1 GB | |
| 45.5 GB | 47.5 GB | 3-bit small — 3-bit small quantization | Does not fit: needs 46.4 GB | |
| 50.6 GB | 52.6 GB | 3-bit medium — 3-bit medium quantization | Does not fit: needs 51.5 GB | |
| 62.2 GB | 64.2 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Does not fit: needs 63.1 GB | |
| 73.7 GB | 75.8 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Does not fit: needs 74.7 GB | |
| 85.3 GB | 87.3 GB | 6-bit, very good — 6-bit quantization, very good quality | Does not fit: needs 86.3 GB | |
| 103.3 GB | 105.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 104.3 GB |
Selected Q4_K_M
4-bit medium quantization63.1GBat 16K context
- Weights
- 61.7 GB
- KV cache
- 1.1 GB
- Overhead
- 0.3 GB
Doesn't fit RTX 4090 (reference) · needs 39.1 GB more
Or .
Which Hardware Can Run Ring Flash 2.0?
Q4_K_M · 62.2 GBRing Flash 2.0 (Q4_K_M) requires 62.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 81+ GB is recommended. Using the full 33K context window can add up to 2.0 GB, bringing total usage to 64.2 GB. No consumer GPU has enough memory. 31 devices with unified memory can run Ring Flash 2.0, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).
32–40 GB
Q4_K_M · best fit IQ2_XXS (29.7 GB)
1 GPU · 2 Mac classes · 1 device family
RTX 5090~309Simulate output on RTX 5090M5 Max36 GB~85Simulate output on Mac Studio M5 Max (36 GB)M3 Pro*36 GB~44Simulate output on MacBook Pro 14" M3 Pro (36 GB)Orin~68Simulate output on NVIDIA Jetson AGX Orin 32GB64–80 GB
fits · 0.9 GB free · up to 16K ctx
1 device family
Orin~34Simulate output on NVIDIA Jetson AGX Orin 64GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
How good is Ring Flash 2.0?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
This is the single board showing results for this model; badges show who ran the test.
- LMArena Text
LMArena · Oct 2026Arena
Arena1322.4
#93/188 open
Similar models
5| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Llama 4 Scout 17B 16E InstructMeta | 82.1 | 77 GB | —needs > 24 GB | same size |
| C4ai Command A 03 2025Cohere | 79.8 | 78.7 GB | —needs > 24 GB | same size |
| GLM 4.6VZ.ai | — | 66 GB | —needs > 24 GB | same size |
| C4ai Command R Plus 08 2024Cohere | — | 73.8 GB | —needs > 24 GB | same size |
| Solar Open 100BUpstage | — | 63.5 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- Inclusion AI
- Parameters
- 102.9B
- Active parameters
- 6.2B per token
- Architecture
- BailingMoeV2ForCausalLM
- Context Length
- 32,768 tokens
- Vocabulary Size
- 157,184
- Release Date
- 2025-09-19
- License
- MIT
Frequently Asked Questions
How much VRAM does Ring Flash 2.0 need?
Ring Flash 2.0 requires 62.2 GB of VRAM at Q4_K_M, or 206.2 GB at BF16. Full 33K context adds up to 2.0 GB (64.2 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 102.9B × 4.8 bits ÷ 8 = 61.7 GB
KV Cache + Overhead ≈ 0.5 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 ≈ 2.5 GB (at full 33K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run Ring Flash 2.0?
Yes, at IQ2_XS (31.3 GB) or lower. Higher quantizations like IQ2_S (32.6 GB) exceed the NVIDIA GeForce RTX 5090's 32 GB.
What's the best quantization for Ring Flash 2.0?
For Ring Flash 2.0, Q4_K_M (62.2 GB) offers the best balance of quality and VRAM usage. Q5_K_M (73.7 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 28.7 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Ring Flash 2.0 on a Mac?
Yes — Mac Studio M5 Max (36 GB) and 65 other Macs can run Ring Flash 2.0, the smallest at IQ2_XS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 62.2 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 96GB).
Can I run Ring Flash 2.0 locally?
Yes — Ring Flash 2.0 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Ring Flash 2.0?
At Q4_K_M, Ring Flash 2.0 can reach ~139 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 → 3.7 GB active ÷ (8000 × 0.65) = 0.715 ms, plus 32 layers × 0.055 ms = 1.760 ms, so 1000 ÷ 2.475 ms = ~404 tok/s
Estimated speed at Q4_K_M (62.2 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Ring Flash 2.0?
At Q4_K_M, the download is about 61.73 GB. The full-precision BF16 version is 205.78 GB. The smallest option (IQ2_XXS) is 28.29 GB.
Which GPUs can run Ring Flash 2.0?
No single consumer GPU has enough VRAM to run Ring Flash 2.0 at Q4_K_M (62.2 GB). Multi-GPU or professional hardware is required.
Which devices can run Ring Flash 2.0?
31 devices with unified memory can run Ring Flash 2.0 at Q4_K_M (62.2 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.