Deep ReinforceOrnithQwen3_5MoeForConditionalGeneration

Ornith 1.0 397B — Hardware Requirements & GPU Compatibility

ChatFunctions

Ornith-1.0-397B is a Mixture-of-Experts model from Deep Reinforce, the largest member of the original Ornith family, built for agentic coding. It was post-trained on top of Google's Gemma 4 and Alibaba's Qwen3.5 using a self-improving reinforcement-learning framework that learns to generate not just solution rollouts but the scaffolding that drives them, aiming for better search trajectories on tasks like terminal use and repository-scale software engineering. It supports a 262,144-token context window and is released under the MIT license, described by its publisher as globally accessible with no regional limits. At roughly 397 billion parameters, 4-bit quantization needs around 228GB of memory, putting local use in server or multi-GPU territory.

396.8B params · MoE18B active262K contextMITJun 2026170.7K downloads261 likes3.3K quant downloads

Your hardware

Verdicts below are for the reference RTX 4090. Set yours to personalise the page.

Detection reads your browser's GPU name with no data sent anywhere.

Ornith 1.0 397B runs on a 256 GB+ Mac or a datacenter GPU

Best local pick → Mac Studio (M3 Ultra, 256GB) · Q4_K_M · 238.6 GB · ~33 tok/s

Macs from 128 GB (IQ2_XS · ~46 tok/s)

MIT

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Ornith 1.0 397B

One recommended way. Every other method, OS and quant is in the step-by-step guide.

Other ways

each opens step-by-step

Run in cloud

Fits on B300 (288 GB) (49 GB headroom) · Q4_K_M

Generation speed: ~187 tok/s; generation speed
Cost per 1M output tokens: $10.33; per 1M output tokens

Deployon RunPodBest optionfrom $6.94/hr

Compare GPUs →

Run Ornith 1.0 397B

238.6 GB VRAM at 16K context · runs on one Instinct MI325X 256 GB

Works on macOS, Windows and Linux.

Install Ollama (Linux)
$curl -fsSL https://ollama.com/install.sh | sh
Install Ollama (macOS / Windows)
Download and run the app; it installs the ollama command.
Download Ollama ↗
Run Q4_K_M
$ollama run hf.co/unsloth/Ornith-1.0-397B-GGUF:Q4_K_M
Raise the context window
Type it inside the chat. For the API server instead, start it with OLLAMA_CONTEXT_LENGTH=16384 ollama serve.
$/set parameter num_ctx 16384
A larger context needs more memory than the weights alone.
Use the API
Ollama serves its API on http://localhost:11434.

Works on macOS, Windows and Linux.

Install LM Studio
Desktop app for macOS (Apple silicon), Windows and Linux.
lmstudio.ai ↗
Download the model
In the app press Ctrl/Cmd + Shift + M and paste https://huggingface.co/unsloth/Ornith-1.0-397B-GGUF, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/unsloth/Ornith-1.0-397B-GGUF@Q4_K_M
Open on Hugging Face ↗
Load and chat
Open the model loader, select the model and set Context Length to 16384. Leave GPU offload at maximum if it fits.
Start the local server (optional)
$lms server start

Works on macOS, Windows and Linux.

Install llama.cpp (macOS)
MacPorts, Nix and conda-forge also work.
$brew install llama.cpp
Install options ↗
Install llama.cpp (Windows)
$winget install llama.cpp
Install llama.cpp (Linux)
Download the prebuilt llama-*-bin-ubuntu-* archive for your GPU backend (CPU, Vulkan) from the releases page, unpack it and run the binaries from that folder. Homebrew (brew install llama.cpp) and conda-forge also work on Linux.
llama.cpp releases ↗
Start the server
Downloads the model on first run. The chat UI and an OpenAI-compatible API are at http://localhost:8080.
$llama-server -hf unsloth/Ornith-1.0-397B-GGUF:Q4_K_M -c 16384 -ngl 99 --port 8080

Works on macOS, Windows and Linux.

Option A: Docker Model Runner
Docker Desktop: Settings > AI > Enable Docker Model Runner. Docker Engine on Linux: install the docker-model-plugin package.
$docker model run hf.co/unsloth/Ornith-1.0-397B-GGUF:Q4_K_M
Model Runner docs ↗
The quant tag after the repo is how Hugging Face cards write it; if your Docker version rejects it, drop the tag and pick the quant by repo.
Option B: llama.cpp server image (NVIDIA)
Needs the NVIDIA Container Toolkit. Models are cached in the llama-cache volume. Open http://localhost:8080.
$docker run --gpus all -p 8080:8080 -v llama-cache:/root/.cache ghcr.io/ggml-org/llama.cpp:server-cuda -hf unsloth/Ornith-1.0-397B-GGUF:Q4_K_M -c 16384 -ngl 99 --host 0.0.0.0 --port 8080
Image tags (cuda13, vulkan, rocm) ↗
Cache path inside the image is not in the docs; remove the -v flag if the volume complains.

Linux only.

Install vLLM
Linux only; on Windows run it inside WSL2. NVIDIA GPU shown; AMD ROCm, Intel and CPU builds have their own install pages.
$uv venv --python 3.12 --seed
$source .venv/bin/activate
$uv pip install vllm --torch-backend=auto
vLLM installation ↗
Serve the model
OpenAI-compatible API at http://localhost:8000/v1. Splits the model over 8 GPUs of one server (tensor parallel).
$vllm serve deepreinforce-ai/Ornith-1.0-397B --max-model-len 16384 --tensor-parallel-size 8
Runs the original weights, not the Q4_K_M GGUF; needs ~873 GB of GPU memory.

macOS only.

MLX supports only macOS, so follow these steps for that setup.

Install mlx-lm
$pip install mlx-lm
Chat in the terminal
$mlx_lm.chat --model pipenetwork/Ornith-1.0-397B-mlx-4bit
Downloads the model on first run. The MLX repo has its own quantisation; it may not match the quant selected above.
Or start a local server
Listens on http://localhost:8080.
$mlx_lm.server --model pipenetwork/Ornith-1.0-397B-mlx-4bit

Works on macOS, Windows and Linux.

Install Jan
jan.ai ↗
Find the model
Open the Hub tab, paste this model ID in Search, and choose Q4_K_M if offered.
$unsloth/Ornith-1.0-397B-GGUF
Use the local API (optional)
Settings > Local API Server > Start Server. It listens on http://127.0.0.1:1337.

Rent GPUs hourly to run Q4_K_M right there. Pricing comes from partner marketplaces and varies.

Available GPU rentals for Q4_K_M, prioritized by value
GPU	Speed	$/1M tokens	Rent
B300 (288 GB)CheapestCheapest49 GB free	~187 tok/s	$10.33	Deploy on RunPodfrom $6.94/hr (opens in new tab)	288 GB · 49 GB free · ~187 tok/s · $10.33 / 1M tokens

Compare GPUs →

How Much VRAM Does Ornith 1.0 397B Need?

Select a quantization level; the listed hardware adheres to it.

VRAM	+32K	+128K	File	Quality	Runs on
109.5 GB	109.9 GB	111.4 GB	124.2 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Runs on Instinct MI250X 128 GB
119.5 GB	119.9 GB	121.4 GB	118.5 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Runs on Instinct MI250X 128 GB
124.4 GB	124.8 GB	126.3 GB	120.5 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Runs on Instinct MI250X 128 GB
134.3 GB	134.7 GB	136.2 GB	124.4 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Runs on H200 NVL 141 GB
154.2 GB	154.6 GB	156.1 GB	147.6 GB	3-bit · i-quant — Importance-weighted 3-bit	Runs on B200 192 GB
164.1 GB	164.5 GB	166 GB	181.4 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Runs on B200 192 GB
169.1 GB	169.5 GB	171 GB	161.9 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Runs on B200 192 GB
169.1 GB	169.5 GB	171 GB	139.5 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Runs on B200 192 GB
174.0 GB	174.4 GB	175.9 GB	172.9 GB	3-bit small — 3-bit small quantization	Runs on B200 192 GB
179.0 GB	179.4 GB	180.9 GB	189.6 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Runs on B200 192 GB
193.9 GB	194.3 GB	195.8 GB	180 GB	3-bit medium — 3-bit medium quantization	Runs on Instinct MI325X 256 GB
198.8 GB	199.2 GB	200.7 GB	225.4 GB	4-bit · legacy — 4-bit legacy quantization	Runs on Instinct MI325X 256 GB
203.8 GB	204.2 GB	205.7 GB	189.5 GB	3-bit large — 3-bit large quantization	Runs on Instinct MI325X 256 GB
213.7 GB	214.1 GB	215.6 GB	192.1 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Runs on Instinct MI325X 256 GB
223.6 GB	224 GB	225.5 GB	249.1 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Runs on Instinct MI325X 256 GB
223.6 GB	224 GB	225.5 GB	229.7 GB	4-bit small — 4-bit small quantization	Runs on Instinct MI325X 256 GB
223.6 GB	224 GB	225.5 GB	196.1 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Runs on Instinct MI325X 256 GB
238.5 GB	238.9 GB	240.4 GB	245 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Runs on Instinct MI325X 256 GB
243.5 GB	243.9 GB	245.4 GB	242.6 GB	4-bit large — 4-bit large quantization	Runs on Instinct MI325X 256 GB
273.2 GB	273.6 GB	275.1 GB	277.5 GB	5-bit small — 5-bit small quantization	Runs on Instinct MI350X 288 GB
283.1 GB	283.5 GB	285 GB	295.3 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Runs on Instinct MI350X 288 GB
327.8 GB	328.2 GB	329.7 GB	329.5 GB	6-bit, very good — 6-bit quantization, very good quality	Runs on 5× H100 SXM 80 GB
397.2 GB	397.6 GB	399.1 GB	421.5 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Runs on 6× H100 SXM 80 GB
794.0 GB	794.4 GB	795.9 GB	793 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Runs on 7× H200 SXM 141 GB

Selected Q4_K_M

4-bit medium quantization

238.6GBat 16K context

Weights: 238.1 GB
KV cache: 0.3 GB
Overhead: 0.3 GB

Server-class: ~238.6 GB at 16K context · one Instinct MI325X 256 GB

Or rent B300 from $6.94/hr.

Which Hardware Can Run Ornith 1.0 397B?

Q4_K_M · 238.5 GB

Ornith 1.0 397B (Q4_K_M) requires 238.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 311+ GB is recommended. Using the full 262K context window can add up to 3.9 GB, bringing total usage to 242.4 GB. No consumer GPU has enough memory. 2 devices with unified memory can run Ornith 1.0 397B, including Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 256GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Consumer GPUs (up to 32 GB) and Macs up to 96 GB: too small for this model (needs ≥ 109.7 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →

Alternatives to Ornith 1.0 397B

Ornith family · 11 models →

Similar models

Model	Score	VRAM	tok/s	Why it is similar
Qwen3.5 397B A17BAlibaba	113.0	242.6 GB	—needs > 24 GB	same size
MiniMax M3MiniMax	112.5	258 GB	—needs > 24 GB	same size
Llama 4 Maverick 17B 128E InstructMeta	87.1	272.7 GB	—needs > 24 GB	same size
Llama 3.1 405B InstructMeta	83.2	275.6 GB	—needs > 24 GB	same size
Intern S2 397BInternLM	—	242.6 GB	—needs > 24 GB	same size
Trinity Large ThinkingArcee AI	—	799.6 GB	—needs > 24 GB	same size

Specifications

Publisher: Deep Reinforce
Family: Ornith
Parameters: 396.8B
Active parameters: 17.8B per token
Architecture: Qwen3_5MoeForConditionalGeneration
Context Length: 262,144 tokens
Vocabulary Size: 248,320
Release Date: 2026-06-23
License: MIT

Resources

Frequently Asked Questions

How much VRAM does Ornith 1.0 397B need?

Ornith 1.0 397B requires 238.5 GB of VRAM at Q4_K_M, or 794.0 GB at BF16. Full 262K context adds up to 3.9 GB (242.4 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 396.8B × 4.8 bits ÷ 8 = 238.1 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 ≈ 4.3 GB (at full 262K context)

VRAM usage by quantization

Q4_K_M

238.5 GB

Q4_K_M + full context

242.4 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run Ornith 1.0 397B?

No — Ornith 1.0 397B requires at least 109.5 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.

What's the best quantization for Ornith 1.0 397B?

For Ornith 1.0 397B, Q4_K_M (238.5 GB) offers the best balance of quality and VRAM usage. Q4_K_L (243.5 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 109.5 GB.

VRAM requirement by quantization

IQ2_XXS

109.5 GB~53%

IQ3_S

169.1 GB~75%

Q3_K_L

203.8 GB~86%

Q4_K_M ★

238.5 GB~89%

Q4_K_L

243.5 GB~90%

BF16

794.0 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Ornith 1.0 397B on a Mac?

Yes — Mac Pro M2 Ultra (128 GB) and 14 other Macs can run Ornith 1.0 397B, 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 238.5 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 256GB).

Can I run Ornith 1.0 397B locally?

Yes — Ornith 1.0 397B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Ornith 1.0 397B?

At Q4_K_M, Ornith 1.0 397B can reach ~70 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 B300 → 10.7 GB active ÷ (8000 × 0.65) = 2.06 ms, plus 60 layers × 0.055 ms = 3.30 ms, so 1000 ÷ 5.36 ms = ~187 tok/s

Estimated speed at Q4_K_M (238.5 GB)

NVIDIA B300

~187 tok/s

AMD Instinct MI350X

~70 tok/s

AMD Instinct MI355X

~70 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of Ornith 1.0 397B?

At Q4_K_M, the download is about 238.08 GB. The full-precision BF16 version is 793.60 GB. The smallest option (IQ2_XXS) is 109.12 GB.

Which GPUs can run Ornith 1.0 397B?

No single consumer GPU has enough VRAM to run Ornith 1.0 397B at Q4_K_M (238.5 GB). Multi-GPU or professional hardware is required.

Which devices can run Ornith 1.0 397B?

2 devices with unified memory can run Ornith 1.0 397B at Q4_K_M (238.5 GB), including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.