Ring 2.6 1T — Hardware Requirements & GPU Compatibility
ChatFunctions
Ring-2.6-1T is Inclusion AI's trillion-parameter mixture-of-experts reasoning model, aimed at agentic workflows rather than plain question answering: multi-step task execution, tool invocation, contextual planning, and long-horizon stability in enterprise and coding scenarios. It exposes a "Reasoning Effort" control with high and xhigh settings, letting developers trade latency for reasoning depth, and was trained with an asynchronous reinforcement-learning architecture combined with the IcePop algorithm, carried over from its predecessor Ring-1T, to stabilize RL training at trillion-parameter scale. With roughly 1 trillion total and 64.5 billion active parameters per token, it needs a multi-GPU server even once quantized. Context length is 131,072 tokens natively, extendable to 262,144 tokens with YaRN scaling. It is released under the MIT license, permitting unrestricted commercial and research use, and was published in May 2026, as a successor to Ring-1T.
Ring 2.6 1T runs on a GPU cluster
BF16 needs ~2.1 TB · about 4 nodes of 8× H100 SXM 80 GB
MIT
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Ring 2.6 1T
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve inclusionAI/Ring-2.6-1T --max-model-len 16384 --tensor-parallel-size 8 --pipeline-parallel-size <nodes>Size the nodes from the table above
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Rent 4× 8-GPU H100 SXM nodes · BF16
- Estimated hourly price
- ~$86.08/hr
- est., whole cluster
- GPUs
- 32 GPUs
- 2.6 TB of VRAM
How Much VRAM Does Ring 2.6 1T Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 2056.5 GB | 2054.2 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on 4 nodes of 8× H100 SXM |
Selected BF16
Brain floating point 162052.6GBat 16K context
- Weights
- 2050.8 GB
- KV cache
- 1.5 GB
- Overhead
- 0.3 GB
Server-class: ~2.1 TB at 16K context · about 4 nodes of 8× H100 SXM 80 GB
Which Hardware Can Run Ring 2.6 1T?
BF16 · 2056.5 GBRing 2.6 1T (BF16) requires 2056.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 2674+ GB is recommended. Using the full 131K context window can add up to 6.7 GB, bringing total usage to 2063.2 GB. No consumer GPU has enough memory.
Needs server-class hardware
Ring 2.6 1T runs on a GPU cluster · BF16 needs ~2.1 TB
| Hardware | Quant | Memory | Context | Hardware page | |
|---|---|---|---|---|---|
| GPU cluster · 4 nodes of 8× H100 SXM 80 GB | BF16 | 2052.6 GB | 128K ctx | H100 SXM → | BF16 · 2052.6 GB · 128K ctx |
By VRAM tier · typical GPUs and Macs at BF16
Consumer GPUs (up to 32 GB) and Macs up to 512 GB: too small for this model (needs ≥ 2.1 TB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
How good is Ring 2.6 1T?
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
Here are all 3 boards that have results for this model, sorted by their weight in the llmrun Score; badges show who ran the test.
Similar models
4| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Kimi K2.6Moonshot AI | 120.9 | 617.6 GB | —needs > 24 GB | same size |
| MiMo V2.5 ProXiaomi | 116.4 | 616 GB | —needs > 24 GB | same size |
| InklingThinking Machines | 116.0 | 575.1 GB | —needs > 24 GB | same size |
| MiMo V2.6 Pro MOPDXiaomi | — | 614.1 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- Inclusion AI
- Family
- Ring
- Parameters
- 1025.4B
- Active parameters
- 64.3B per token
- Architecture
- BailingMoeV2_5ForCausalLM
- Context Length
- 131,072 tokens
- Vocabulary Size
- 157,184
- Release Date
- 2026-05-14
- License
- MIT
Frequently Asked Questions
How much VRAM does Ring 2.6 1T need?
Ring 2.6 1T requires 2056.5 GB of VRAM at BF16. Full 131K context adds up to 6.7 GB (2063.2 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 1025.4B × 16 bits ÷ 8 = 2050.8 GB
KV Cache + Overhead ≈ 5.7 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 ≈ 12.4 GB (at full 131K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run Ring 2.6 1T?
No — Ring 2.6 1T requires at least 2056.5 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
Can I run Ring 2.6 1T on a Mac?
Ring 2.6 1T requires at least 2056.5 GB at BF16. None of the 146 Macs we list has that much usable memory (the largest has 508.5 GB after macOS overhead).
Can I run Ring 2.6 1T locally?
Yes — Ring 2.6 1T can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
What's the download size of Ring 2.6 1T?
At BF16, the download is about 2050.83 GB.
Which GPUs can run Ring 2.6 1T?
No single consumer GPU has enough VRAM to run Ring 2.6 1T at BF16 (2056.5 GB). Multi-GPU or professional hardware is required.
Which devices can run Ring 2.6 1T?
5 GB).